diff --git a/2026_tdl_challenge/analysis_gread.ipynb b/2026_tdl_challenge/analysis_gread.ipynb new file mode 100644 index 000000000..23b0ac358 --- /dev/null +++ b/2026_tdl_challenge/analysis_gread.ipynb @@ -0,0 +1,780 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "cell-01", + "metadata": {}, + "source": [ + "# GREAD on GraphUniverse — supplementary transfer analysis\n", + "\n", + "Model: GREAD (Choi et al., ICML 2023, [arXiv:2211.14208](https://arxiv.org/abs/2211.14208)), config `graph/gread` (`beta_diag: true`, the per-channel reaction gate\n", + "used by every tuned setting of the reference implementation).\n", + "\n", + "This notebook is **supplementary** to the official `run_evaluation.ipynb`. It\n", + "re-trains nothing: it reads the committed `results.json` from the official grid\n", + "and analyses the part of it that the official artifacts summarise only\n", + "partially — the **out-of-distribution (OOD) evaluations**.\n", + "\n", + "The official run trains on each of the 12 GraphUniverse settings and evaluates\n", + "every checkpoint on all 12, so each model yields a full 12x12 transfer matrix\n", + "per task, per seed: 432 measurements per task, of which only the 36 diagonal\n", + "ones appear in the main heatmaps.\n", + "\n", + "We use them to ask three questions the challenge itself poses:\n", + "\n", + "1. Is performance governed by the distribution a model is **evaluated** on, or\n", + " by the **mismatch** between training and evaluation distributions?\n", + "2. Is transfer across homophily levels **symmetric**?\n", + "3. Which structural axis — homophily, average degree, or power-law exponent —\n", + " dominates?\n", + "\n", + "The headline result is that the two tasks answer question 1 in **opposite**\n", + "ways." + ] + }, + { + "cell_type": "code", + "execution_count": 1, + "id": "cell-02", + "metadata": { + "execution": { + "iopub.execute_input": "2026-08-01T20:07:36.727803Z", + "iopub.status.busy": "2026-08-01T20:07:36.727715Z", + "iopub.status.idle": "2026-08-01T20:07:37.346509Z", + "shell.execute_reply": "2026-08-01T20:07:37.345700Z" + } + }, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "864 measurements (72 in-distribution, 792 OOD)\n", + "\n", + "triangle-counting MSE/triangles spans 0.01082 to 2164 -> analysed as log10\n" + ] + }, + { + "data": { + "text/html": [ + "
\n", + "\n", + "\n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + "
tasktrainevalseedin_distributionscoretrain_homophilytrain_degreetrain_powerlaweval_homophilyeval_degreeeval_powerlawvalue
0community_detectionh_lo__d_lo__pl_loh_lo__d_lo__pl_lo42True0.338834lolololololo0.338834
1community_detectionh_lo__d_lo__pl_loh_lo__d_lo__pl_hi42False0.330354lololololohi0.330354
2community_detectionh_lo__d_lo__pl_loh_lo__d_hi__pl_lo42False0.371724lolololohilo0.371724
3community_detectionh_lo__d_lo__pl_loh_lo__d_hi__pl_hi42False0.367255lolololohihi0.367255
4community_detectionh_lo__d_lo__pl_loh_mid__d_lo__pl_lo42False0.350371lololomidlolo0.350371
\n", + "
" + ], + "text/plain": [ + " task train eval seed \\\n", + "0 community_detection h_lo__d_lo__pl_lo h_lo__d_lo__pl_lo 42 \n", + "1 community_detection h_lo__d_lo__pl_lo h_lo__d_lo__pl_hi 42 \n", + "2 community_detection h_lo__d_lo__pl_lo h_lo__d_hi__pl_lo 42 \n", + "3 community_detection h_lo__d_lo__pl_lo h_lo__d_hi__pl_hi 42 \n", + "4 community_detection h_lo__d_lo__pl_lo h_mid__d_lo__pl_lo 42 \n", + "\n", + " in_distribution score train_homophily train_degree train_powerlaw \\\n", + "0 True 0.338834 lo lo lo \n", + "1 False 0.330354 lo lo lo \n", + "2 False 0.371724 lo lo lo \n", + "3 False 0.367255 lo lo lo \n", + "4 False 0.350371 lo lo lo \n", + "\n", + " eval_homophily eval_degree eval_powerlaw value \n", + "0 lo lo lo 0.338834 \n", + "1 lo lo hi 0.330354 \n", + "2 lo hi lo 0.371724 \n", + "3 lo hi hi 0.367255 \n", + "4 mid lo lo 0.350371 " + ] + }, + "execution_count": 1, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "import json\n", + "from pathlib import Path\n", + "\n", + "import matplotlib.pyplot as plt\n", + "import numpy as np\n", + "import pandas as pd\n", + "\n", + "STUDY_ID = \"2026-08-02_gread-vc\"\n", + "RESULTS = Path(\"outputs\") / STUDY_ID / \"results.json\"\n", + "if not RESULTS.exists(): # allow running from the repository root\n", + " RESULTS = Path(\"2026_tdl_challenge\") / RESULTS\n", + "\n", + "# metric key, display label, better direction, whether to analyse on a log scale\n", + "TASK_METRIC = {\n", + " \"community_detection\": (\"test_best_rerun_accuracy\", \"accuracy\", \"max\", False),\n", + " \"triangle_counting\": (\"test_mse_by_total_triangles\", \"MSE / triangles\", \"min\", True),\n", + "}\n", + "\n", + "REGIME_ORDER = [\n", + " f\"h_{h}__d_{d}__pl_{p}\"\n", + " for h in (\"lo\", \"mid\", \"hi\")\n", + " for d in (\"lo\", \"hi\")\n", + " for p in (\"lo\", \"hi\")\n", + "]\n", + "HOM_ORDER = [\"lo\", \"mid\", \"hi\"]\n", + "\n", + "\n", + "def load_long(path):\n", + " \"\"\"Flatten results.json into one row per (task, train regime, eval regime, seed).\"\"\"\n", + " payload = json.loads(Path(path).read_text())\n", + " rows = []\n", + " for run in payload[\"results\"]:\n", + " task = run[\"experiment\"]\n", + " metric = TASK_METRIC[task][0]\n", + " train = run[\"run_slug\"]\n", + " rows.append(\n", + " dict(task=task, train=train, eval=train, seed=run[\"train_seed\"],\n", + " in_distribution=True, score=run[metric])\n", + " )\n", + " for eval_slug, metrics in (run.get(\"ood_test\") or {}).items():\n", + " rows.append(\n", + " dict(task=task, train=train, eval=eval_slug, seed=run[\"train_seed\"],\n", + " in_distribution=False, score=metrics[metric])\n", + " )\n", + " frame = pd.DataFrame(rows)\n", + " for col in (\"train\", \"eval\"):\n", + " parts = frame[col].str.split(\"__\", expand=True)\n", + " frame[f\"{col}_homophily\"] = parts[0].str.removeprefix(\"h_\")\n", + " frame[f\"{col}_degree\"] = parts[1].str.removeprefix(\"d_\")\n", + " frame[f\"{col}_powerlaw\"] = parts[2].str.removeprefix(\"pl_\")\n", + " # Analysis scale: triangle-counting errors span several orders of magnitude,\n", + " # so a raw-scale summary would be dominated by a handful of extreme cells.\n", + " frame[\"value\"] = frame.score\n", + " log_tasks = [t for t, spec in TASK_METRIC.items() if spec[3]]\n", + " mask = frame.task.isin(log_tasks)\n", + " frame.loc[mask, \"value\"] = np.log10(frame.loc[mask, \"score\"].clip(lower=1e-6))\n", + " return frame\n", + "\n", + "\n", + "df = load_long(RESULTS)\n", + "print(f\"{len(df)} measurements \"\n", + " f\"({df.in_distribution.sum()} in-distribution, {(~df.in_distribution).sum()} OOD)\")\n", + "print(\"\\ntriangle-counting MSE/triangles spans \"\n", + " f\"{df[df.task == 'triangle_counting'].score.min():.4g} to \"\n", + " f\"{df[df.task == 'triangle_counting'].score.max():.4g} \"\n", + " \"-> analysed as log10\")\n", + "df.head()" + ] + }, + { + "cell_type": "markdown", + "id": "cell-03", + "metadata": {}, + "source": [ + "## 1. Evaluation distribution, or train/eval mismatch?\n", + "\n", + "An OOD drop is usually read as a transfer failure. But the same drop appears if\n", + "the target distribution is simply harder, wherever the model was trained. The\n", + "two explanations make different predictions:\n", + "\n", + "- *Mismatch matters*: scores depend on the **training** regime.\n", + "- *Target difficulty matters*: scores depend on the **evaluation** regime.\n", + "\n", + "We quantify this as the fraction of score variance explained by grouping the\n", + "432 measurements by evaluation regime, versus by training regime." + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "id": "cell-04", + "metadata": { + "execution": { + "iopub.execute_input": "2026-08-01T20:07:37.349490Z", + "iopub.status.busy": "2026-08-01T20:07:37.349051Z", + "iopub.status.idle": "2026-08-01T20:07:37.367538Z", + "shell.execute_reply": "2026-08-01T20:07:37.366693Z" + } + }, + "outputs": [ + { + "data": { + "text/html": [ + "
\n", + "\n", + "\n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + "
analysedby_eval_regimeby_train_regime
task
community_detectionaccuracy0.8500.030
triangle_countinglog10(MSE / triangles)0.1160.407
\n", + "
" + ], + "text/plain": [ + " analysed by_eval_regime by_train_regime\n", + "task \n", + "community_detection accuracy 0.850 0.030\n", + "triangle_counting log10(MSE / triangles) 0.116 0.407" + ] + }, + "execution_count": 2, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "def variance_explained(frame, by, col=\"value\"):\n", + " \"\"\"Fraction of variance in `col` explained by grouping on `by`.\"\"\"\n", + " group_mean = frame.groupby(by)[col].transform(\"mean\")\n", + " resid = ((frame[col] - group_mean) ** 2).mean()\n", + " total = ((frame[col] - frame[col].mean()) ** 2).mean()\n", + " return 1.0 - resid / total\n", + "\n", + "\n", + "summary = []\n", + "for task, sub in df.groupby(\"task\"):\n", + " metric, label, _, is_log = TASK_METRIC[task]\n", + " summary.append(\n", + " dict(\n", + " task=task,\n", + " analysed=f\"log10({label})\" if is_log else label,\n", + " by_eval_regime=variance_explained(sub, \"eval\"),\n", + " by_train_regime=variance_explained(sub, \"train\"),\n", + " )\n", + " )\n", + "variance_table = pd.DataFrame(summary).set_index(\"task\").round(3)\n", + "variance_table" + ] + }, + { + "cell_type": "markdown", + "id": "cell-05", + "metadata": {}, + "source": [ + "The two tasks behave in opposite ways, and this is the main result of the\n", + "notebook.\n", + "\n", + "**Community detection** is governed by the evaluation regime: it explains an\n", + "order of magnitude more variance than the training regime. The model transfers\n", + "well; the regimes differ in intrinsic difficulty. Reporting an \"OOD gap\" here\n", + "without this decomposition would attribute to distribution shift what is mostly\n", + "a difficulty effect.\n", + "\n", + "**Triangle counting** is the reverse: the training regime dominates. This is a\n", + "genuine distribution-shift failure. A count is an extensive quantity whose\n", + "typical magnitude changes by orders of magnitude across regimes, so a regressor\n", + "calibrated on one regime systematically mispredicts the scale elsewhere — a\n", + "calibration failure rather than a failure to perceive structure." + ] + }, + { + "cell_type": "code", + "execution_count": 3, + "id": "cell-06", + "metadata": { + "execution": { + "iopub.execute_input": "2026-08-01T20:07:37.369619Z", + "iopub.status.busy": "2026-08-01T20:07:37.369504Z", + "iopub.status.idle": "2026-08-01T20:07:37.801557Z", + "shell.execute_reply": "2026-08-01T20:07:37.800784Z" + } + }, + "outputs": [ + { + "data": { + "image/png": "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", + "text/plain": [ + "
" + ] + }, + "metadata": {}, + "output_type": "display_data" + } + ], + "source": [ + "fig, axes = plt.subplots(1, 2, figsize=(18, 7.5))\n", + "for ax, (task, (metric, label, direction, is_log)) in zip(axes, TASK_METRIC.items()):\n", + " sub = df[df.task == task]\n", + " matrix = (\n", + " sub.pivot_table(index=\"train\", columns=\"eval\", values=\"value\", aggfunc=\"mean\")\n", + " .reindex(index=REGIME_ORDER, columns=REGIME_ORDER)\n", + " )\n", + " data = matrix.to_numpy()\n", + " im = ax.imshow(data, cmap=\"RdBu_r\" if direction == \"max\" else \"RdBu\")\n", + " ax.set_xticks(range(len(REGIME_ORDER)))\n", + " ax.set_yticks(range(len(REGIME_ORDER)))\n", + " ax.set_xticklabels(REGIME_ORDER, rotation=90, fontsize=7)\n", + " ax.set_yticklabels(REGIME_ORDER, fontsize=7)\n", + " ax.set_xlabel(\"evaluation regime\")\n", + " ax.set_ylabel(\"training regime\")\n", + " shown = f\"log10({label})\" if is_log else label\n", + " ax.set_title(f\"{task}\\n{shown} (mean over seeds)\")\n", + " for i in range(data.shape[0]):\n", + " for j in range(data.shape[1]):\n", + " ax.text(j, i, f\"{data[i, j]:.2f}\", ha=\"center\", va=\"center\", fontsize=6)\n", + " fig.colorbar(im, ax=ax, fraction=0.046, label=shown)\n", + "fig.suptitle(\"Full 12x12 transfer matrices (the official heatmaps show only the diagonal)\")\n", + "fig.tight_layout()\n", + "plt.show()" + ] + }, + { + "cell_type": "markdown", + "id": "cell-07", + "metadata": {}, + "source": [ + "Read the two panels by comparing how each varies. Community detection varies\n", + "mostly left-to-right (along the evaluation axis); triangle counting varies\n", + "mostly top-to-bottom (along the training axis).\n", + "\n", + "## 2. Is transfer across homophily symmetric?\n", + "\n", + "Collapsing onto the homophily axis isolates the structural property the\n", + "challenge cares most about. Symmetric transfer would give a symmetric matrix." + ] + }, + { + "cell_type": "code", + "execution_count": 4, + "id": "cell-08", + "metadata": { + "execution": { + "iopub.execute_input": "2026-08-01T20:07:37.803218Z", + "iopub.status.busy": "2026-08-01T20:07:37.803108Z", + "iopub.status.idle": "2026-08-01T20:07:37.814389Z", + "shell.execute_reply": "2026-08-01T20:07:37.813840Z" + } + }, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "\n", + "community_detection — accuracy (rows = trained on, cols = evaluated on)\n", + "eval_homophily lo mid hi\n", + "train_homophily \n", + "lo 0.347 0.351 0.496\n", + "mid 0.299 0.405 0.606\n", + "hi 0.275 0.385 0.666\n", + " trained hi -> evaluated lo : 0.275 (native lo = 0.347)\n", + " trained lo -> evaluated hi : 0.496 (native hi = 0.666)\n", + " best generalist (row mean over all eval regimes): trained on h_hi\n", + "\n", + "triangle_counting — log10(MSE / triangles) (rows = trained on, cols = evaluated on)\n", + "eval_homophily lo mid hi\n", + "train_homophily \n", + "lo -0.557 -0.003 0.434\n", + "mid 0.355 0.269 0.450\n", + "hi 1.130 0.702 0.606\n", + " trained hi -> evaluated lo : 1.130 (native lo = -0.557)\n", + " trained lo -> evaluated hi : 0.434 (native hi = 0.606)\n", + " best generalist (row mean over all eval regimes): trained on h_lo\n" + ] + } + ], + "source": [ + "for task, (metric, label, direction, is_log) in TASK_METRIC.items():\n", + " sub = df[df.task == task]\n", + " shown = f\"log10({label})\" if is_log else label\n", + " hom = (\n", + " sub.pivot_table(index=\"train_homophily\", columns=\"eval_homophily\",\n", + " values=\"value\", aggfunc=\"mean\")\n", + " .reindex(index=HOM_ORDER, columns=HOM_ORDER)\n", + " )\n", + " print(f\"\\n{task} — {shown} (rows = trained on, cols = evaluated on)\")\n", + " print(hom.round(3).to_string())\n", + " print(f\" trained hi -> evaluated lo : {hom.loc['hi', 'lo']:.3f} \"\n", + " f\"(native lo = {hom.loc['lo', 'lo']:.3f})\")\n", + " print(f\" trained lo -> evaluated hi : {hom.loc['lo', 'hi']:.3f} \"\n", + " f\"(native hi = {hom.loc['hi', 'hi']:.3f})\")\n", + " best = hom.mean(axis=1)\n", + " best = best.idxmax() if direction == \"max\" else best.idxmin()\n", + " print(f\" best generalist (row mean over all eval regimes): trained on h_{best}\")" + ] + }, + { + "cell_type": "markdown", + "id": "cell-09", + "metadata": {}, + "source": [ + "Transfer is clearly asymmetric, and the asymmetry runs in opposite directions\n", + "for the two tasks. For community detection the high-homophily model is the\n", + "best generalist (row means 0.442 vs 0.437 for mid and 0.398 for low), even\n", + "though it degrades the most on heterophilic graphs (0.275 vs the native\n", + "0.347); its advantage on homophilous evaluations outweighs that loss.\n", + "For triangle counting the low-homophily model is the safest, because it is\n", + "calibrated to the sparse-triangle end of the scale.\n", + "\n", + "## 3. Which structural axis dominates?\n", + "\n", + "Variance explained by each axis of the *evaluation* regime alone." + ] + }, + { + "cell_type": "code", + "execution_count": 5, + "id": "cell-10", + "metadata": { + "execution": { + "iopub.execute_input": "2026-08-01T20:07:37.816355Z", + "iopub.status.busy": "2026-08-01T20:07:37.816257Z", + "iopub.status.idle": "2026-08-01T20:07:37.831321Z", + "shell.execute_reply": "2026-08-01T20:07:37.830698Z" + } + }, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "axis degree homophily powerlaw\n", + "task side \n", + "community_detection eval 0.025 0.810 0.000\n", + " train 0.002 0.022 0.000\n", + "triangle_counting eval 0.004 0.010 0.000\n", + " train 0.132 0.172 0.054\n", + "\n", + "Mean value by regime level (log10 scale for triangle counting):\n", + " community_detection eval_homophily {'hi': 0.589, 'lo': 0.307, 'mid': 0.381}\n", + " community_detection eval_degree {'hi': 0.447, 'lo': 0.405}\n", + " community_detection eval_powerlaw {'hi': 0.427, 'lo': 0.424}\n", + " triangle_counting train_homophily {'hi': 0.812, 'lo': -0.042, 'mid': 0.358}\n", + " triangle_counting train_degree {'hi': 0.682, 'lo': 0.071}\n", + " triangle_counting train_powerlaw {'hi': 0.18, 'lo': 0.572}\n" + ] + } + ], + "source": [ + "axis_rows = []\n", + "for task, sub in df.groupby(\"task\"):\n", + " for prefix in (\"eval\", \"train\"):\n", + " for axis in (\"homophily\", \"degree\", \"powerlaw\"):\n", + " axis_rows.append(\n", + " dict(task=task, side=prefix, axis=axis,\n", + " variance_explained=variance_explained(sub, f\"{prefix}_{axis}\"))\n", + " )\n", + "axis_table = (\n", + " pd.DataFrame(axis_rows)\n", + " .pivot(index=[\"task\", \"side\"], columns=\"axis\", values=\"variance_explained\")\n", + " .round(3)\n", + ")\n", + "print(axis_table.to_string())\n", + "\n", + "print(\"\\nMean value by regime level (log10 scale for triangle counting):\")\n", + "for task, (metric, label, direction, is_log) in TASK_METRIC.items():\n", + " sub = df[df.task == task]\n", + " side = \"eval\" if task == \"community_detection\" else \"train\"\n", + " for axis in (\"homophily\", \"degree\", \"powerlaw\"):\n", + " means = sub.groupby(f\"{side}_{axis}\").value.mean().round(3).to_dict()\n", + " print(f\" {task:20s} {side}_{axis:10s} {means}\")" + ] + }, + { + "cell_type": "markdown", + "id": "cell-11", + "metadata": {}, + "source": [ + "Homophily dominates the evaluation side of community detection (0.81 vs at\n", + "most 0.03 for any other axis). For triangle counting no single axis\n", + "dominates: on the training side homophily leads (0.17) with average degree\n", + "close behind (0.13), and the power-law exponent is secondary throughout.\n", + "\n", + "## 4. Seed stability\n", + "\n", + "Three seeds per (task, regime). Small dispersion is what licenses reading the\n", + "differences above as real rather than as noise." + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "id": "cell-12", + "metadata": { + "execution": { + "iopub.execute_input": "2026-08-01T20:07:37.833810Z", + "iopub.status.busy": "2026-08-01T20:07:37.833680Z", + "iopub.status.idle": "2026-08-01T20:07:37.841444Z", + "shell.execute_reply": "2026-08-01T20:07:37.841099Z" + } + }, + "outputs": [ + { + "data": { + "text/html": [ + "
\n", + "\n", + "\n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + "
metricmax_seed_stdmedian_seed_stdspread_across_regimes
task
community_detectionaccuracy0.00270.00130.4124
triangle_countingMSE / triangles0.05380.01463.6532
\n", + "
" + ], + "text/plain": [ + " metric max_seed_std median_seed_std \\\n", + "task \n", + "community_detection accuracy 0.0027 0.0013 \n", + "triangle_counting MSE / triangles 0.0538 0.0146 \n", + "\n", + " spread_across_regimes \n", + "task \n", + "community_detection 0.4124 \n", + "triangle_counting 3.6532 " + ] + }, + "execution_count": 6, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "stability = []\n", + "for task, sub in df[df.in_distribution].groupby(\"task\"):\n", + " per_regime = sub.groupby(\"train\").score\n", + " stability.append(\n", + " dict(task=task, metric=TASK_METRIC[task][1],\n", + " max_seed_std=per_regime.std().max(),\n", + " median_seed_std=per_regime.std().median(),\n", + " spread_across_regimes=per_regime.mean().max() - per_regime.mean().min())\n", + " )\n", + "pd.DataFrame(stability).set_index(\"task\").round(4)" + ] + }, + { + "cell_type": "markdown", + "id": "cell-13", + "metadata": {}, + "source": [ + "## Takeaways\n", + "\n", + "1. **The two tasks fail differently.** For community detection, GREAD's score is\n", + " set by the regime it is *evaluated* on (variance explained 0.85)\n", + " and barely by the one it was *trained* on (0.03). The apparent\n", + " OOD penalty is target difficulty, not failed transfer. For triangle counting\n", + " the ordering reverses on a log scale (train 0.41 vs eval\n", + " 0.12): that task does exhibit a real distribution-shift failure.\n", + "2. **Counting fails by miscalibrated scale.** Triangle counts are extensive and\n", + " their magnitude shifts by orders of magnitude across regimes, so a model\n", + " calibrated on one regime mispredicts the scale on another. The worst cells of\n", + " the transfer matrix are exactly the largest scale jumps.\n", + "3. **Homophily transfer is asymmetric**, in opposite directions per task: the\n", + " best community-detection generalist here is the model trained on\n", + " high homophily, whereas the safest triangle counter is the\n", + " one trained on low homophily, which is calibrated to the sparse-triangle end\n", + " of the scale.\n", + "4. **Methodological note.** Any OOD summary of a regression task on this grid\n", + " should be read on a log scale; on the raw scale a few extreme cells dominate\n", + " the statistics and invert the conclusion.\n", + "\n", + "All numbers derive from the committed `results.json`. Nothing here re-trains or\n", + "modifies the official evaluation pipeline." + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/2026_tdl_challenge/outputs/2026-08-02_gread-vc/results.json b/2026_tdl_challenge/outputs/2026-08-02_gread-vc/results.json new file mode 100644 index 000000000..ac1289456 --- /dev/null +++ b/2026_tdl_challenge/outputs/2026-08-02_gread-vc/results.json @@ -0,0 +1,5272 @@ +{ + "metadata": { + "study_id": "2026-08-02_gread-vc", + "model_config": "graph/gread", + "generated_at_utc": "2026-08-01T20:05:11.192867+00:00", + "n_runs": 72, + "train_seeds": [ + 42, + 43, + 44 + ], + "heatmap_note": "Cells show mean \u00b1 std over train_seeds (in-distribution test)." + }, + "results": [ + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0-0.1__deg_1-2.5__gamma_1.5-2__s42", + "train_seed": 42, + "homophily": "h_lo", + "avg_degree": "d_lo", + "power_law": "pl_lo", + "run_slug": "h_lo__d_lo__pl_lo", + "test_loss": 2.12618350982666, + "test_best_rerun_accuracy": 0.33883413672447205, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.3303537368774414, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.37172433733940125, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.36725494265556335, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.35037052631378174, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.34376195073127747, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.3383757472038269, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.3170219361782074, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.43887996673583984, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.4400259852409363, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.4566429853439331, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.45859119296073914, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s42_community_detection__community_detection__00__h_lo__d_lo__pl_lo__s42" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0-0.1__deg_1-2.5__gamma_1.5-2__s43", + "train_seed": 43, + "homophily": "h_lo", + "avg_degree": "d_lo", + "power_law": "pl_lo", + "run_slug": "h_lo__d_lo__pl_lo", + "test_loss": 2.1327407360076904, + "test_best_rerun_accuracy": 0.3361601233482361, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.32874932885169983, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.37229734659194946, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.3653831481933594, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.3535793423652649, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.3428451418876648, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.3394835293292999, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.3188173174858093, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.4389563798904419, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.44036978483200073, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.4598899781703949, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.4634425938129425, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s43_community_detection__community_detection__00__h_lo__d_lo__pl_lo__s43" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0-0.1__deg_1-2.5__gamma_1.5-2__s44", + "train_seed": 44, + "homophily": "h_lo", + "avg_degree": "d_lo", + "power_law": "pl_lo", + "run_slug": "h_lo__d_lo__pl_lo", + "test_loss": 2.135568141937256, + "test_best_rerun_accuracy": 0.33623653650283813, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.3302391469478607, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.36905035376548767, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.36450454592704773, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.3523569405078888, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.34257772564888, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.3391397297382355, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.3185499310493469, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.43712276220321655, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.44201236963272095, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.45809459686279297, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.46374818682670593, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s44_community_detection__community_detection__00__h_lo__d_lo__pl_lo__s44" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0-0.1__deg_1-2.5__gamma_4-5__s42", + "train_seed": 42, + "homophily": "h_lo", + "avg_degree": "d_lo", + "power_law": "pl_hi", + "run_slug": "h_lo__d_lo__pl_hi", + "test_loss": 2.1571204662323, + "test_best_rerun_accuracy": 0.3300863206386566, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.3354725241661072, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.36316755414009094, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.36183053255081177, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.3534265458583832, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.3449079394340515, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.3383757472038269, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.3239743411540985, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.43456336855888367, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.43712276220321655, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.45396897196769714, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.45668119192123413, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s42_community_detection__community_detection__01__h_lo__d_lo__pl_hi__s42" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0-0.1__deg_1-2.5__gamma_4-5__s43", + "train_seed": 43, + "homophily": "h_lo", + "avg_degree": "d_lo", + "power_law": "pl_hi", + "run_slug": "h_lo__d_lo__pl_hi", + "test_loss": 2.149336338043213, + "test_best_rerun_accuracy": 0.330048143863678, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.3353961408138275, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.36427533626556396, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.3617541491985321, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.35205134749412537, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.3424249291419983, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.3371533453464508, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.32240813970565796, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.4278401732444763, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.431774765253067, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.4464435875415802, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.44747498631477356, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s43_community_detection__community_detection__01__h_lo__d_lo__pl_hi__s43" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0-0.1__deg_1-2.5__gamma_4-5__s44", + "train_seed": 44, + "homophily": "h_lo", + "avg_degree": "d_lo", + "power_law": "pl_hi", + "run_slug": "h_lo__d_lo__pl_hi", + "test_loss": 2.157198429107666, + "test_best_rerun_accuracy": 0.32867294549942017, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.3344411253929138, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.36221253871917725, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.36083734035491943, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.3512491285800934, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.34578654170036316, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.3371151387691498, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.32389792799949646, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.4320421814918518, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.4342195689678192, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.4488883912563324, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.45522958040237427, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s44_community_detection__community_detection__01__h_lo__d_lo__pl_hi__s44" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0-0.1__deg_4-5__gamma_1.5-2__s42", + "train_seed": 42, + "homophily": "h_lo", + "avg_degree": "d_hi", + "power_law": "pl_lo", + "run_slug": "h_lo__d_hi__pl_lo", + "test_loss": 1.9829374551773071, + "test_best_rerun_accuracy": 0.3817327618598938, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.3265337347984314, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.3129727244377136, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.3680189549922943, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.36767515540122986, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.35694095492362976, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.3666437566280365, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.3367713391780853, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.5024065971374512, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.5117655992507935, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.5468332171440125, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.5645580291748047, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s42_community_detection__community_detection__02__h_lo__d_hi__pl_lo__s42" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0-0.1__deg_4-5__gamma_1.5-2__s43", + "train_seed": 43, + "homophily": "h_lo", + "avg_degree": "d_hi", + "power_law": "pl_lo", + "run_slug": "h_lo__d_hi__pl_lo", + "test_loss": 1.9749042987823486, + "test_best_rerun_accuracy": 0.381465345621109, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.32714492082595825, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.31530293822288513, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.3708457350730896, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.3721063435077667, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.35812515020370483, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.3664909601211548, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.3397509455680847, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.5073344111442566, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.5170372128486633, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.554969847202301, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.5726182460784912, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s43_community_detection__community_detection__02__h_lo__d_hi__pl_lo__s43" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0-0.1__deg_4-5__gamma_1.5-2__s44", + "train_seed": 44, + "homophily": "h_lo", + "avg_degree": "d_hi", + "power_law": "pl_lo", + "run_slug": "h_lo__d_hi__pl_lo", + "test_loss": 1.9897323846817017, + "test_best_rerun_accuracy": 0.3790205419063568, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.321644127368927, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.31094813346862793, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.3659561574459076, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.3692031502723694, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.3533119559288025, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.3675987422466278, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.34303614497184753, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.5008785724639893, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.5110397934913635, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.5526396036148071, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.5707082152366638, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s44_community_detection__community_detection__02__h_lo__d_hi__pl_lo__s44" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0-0.1__deg_4-5__gamma_4-5__s42", + "train_seed": 42, + "homophily": "h_lo", + "avg_degree": "d_hi", + "power_law": "pl_hi", + "run_slug": "h_lo__d_hi__pl_hi", + "test_loss": 2.026320219039917, + "test_best_rerun_accuracy": 0.36889755725860596, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.3183589279651642, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.30823591351509094, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.37172433733940125, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.37772175669670105, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.3658033609390259, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.3750477433204651, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.35759034752845764, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.5195202231407166, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.5330812335014343, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.5726945996284485, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.6026816368103027, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s42_community_detection__community_detection__03__h_lo__d_hi__pl_hi__s42" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0-0.1__deg_4-5__gamma_4-5__s43", + "train_seed": 43, + "homophily": "h_lo", + "avg_degree": "d_hi", + "power_law": "pl_hi", + "run_slug": "h_lo__d_hi__pl_hi", + "test_loss": 2.0091536045074463, + "test_best_rerun_accuracy": 0.3725265562534332, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.32332491874694824, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.3122851252555847, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.3753533363342285, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.38299334049224854, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.36614716053009033, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.37661394476890564, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.3592711389064789, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.5208954215049744, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.5351822376251221, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.5756742358207703, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.6003132462501526, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s43_community_detection__community_detection__03__h_lo__d_hi__pl_hi__s43" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0-0.1__deg_4-5__gamma_4-5__s44", + "train_seed": 44, + "homophily": "h_lo", + "avg_degree": "d_hi", + "power_law": "pl_hi", + "run_slug": "h_lo__d_hi__pl_hi", + "test_loss": 2.0166783332824707, + "test_best_rerun_accuracy": 0.37256476283073425, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.3214913308620453, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.3109099268913269, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.3748185634613037, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.38123616576194763, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.3663763403892517, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.37474215030670166, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.35720834136009216, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.5196729898452759, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.5318205952644348, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.5702880024909973, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.5955764651298523, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s44_community_detection__community_detection__03__h_lo__d_hi__pl_hi__s44" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0.4-0.6__deg_1-2.5__gamma_1.5-2__s42", + "train_seed": 42, + "homophily": "h_mid", + "avg_degree": "d_lo", + "power_law": "pl_lo", + "run_slug": "h_mid__d_lo__pl_lo", + "test_loss": 1.9429086446762085, + "test_best_rerun_accuracy": 0.4033157527446747, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.2992207109928131, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.2894033193588257, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.3318053185939789, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.33367714285850525, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.38929635286331177, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.4090839624404907, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.40488195419311523, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.5576056241989136, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.5723890066146851, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.6230422258377075, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.6621208786964417, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s42_community_detection__community_detection__04__h_mid__d_lo__pl_lo__s42" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0.4-0.6__deg_1-2.5__gamma_1.5-2__s43", + "train_seed": 43, + "homophily": "h_mid", + "avg_degree": "d_lo", + "power_law": "pl_lo", + "run_slug": "h_mid__d_lo__pl_lo", + "test_loss": 1.9408735036849976, + "test_best_rerun_accuracy": 0.40312474966049194, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.3021621108055115, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.29138970375061035, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.33379173278808594, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.3366185426712036, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.39116814732551575, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.40831995010375977, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.4002215564250946, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.5564213991165161, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.5712048411369324, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.6235770583152771, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.6574986577033997, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s43_community_detection__community_detection__04__h_mid__d_lo__pl_lo__s43" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0.4-0.6__deg_1-2.5__gamma_1.5-2__s44", + "train_seed": 44, + "homophily": "h_mid", + "avg_degree": "d_lo", + "power_law": "pl_lo", + "run_slug": "h_mid__d_lo__pl_lo", + "test_loss": 1.9397659301757812, + "test_best_rerun_accuracy": 0.40461456775665283, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.30021393299102783, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.2915043234825134, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.3329513370990753, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.335778146982193, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.39143556356430054, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.41290396451950073, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.4056841731071472, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.5565360188484192, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.5707846283912659, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.624264657497406, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.6609748601913452, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s44_community_detection__community_detection__04__h_mid__d_lo__pl_lo__s44" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0.4-0.6__deg_1-2.5__gamma_4-5__s42", + "train_seed": 42, + "homophily": "h_mid", + "avg_degree": "d_lo", + "power_law": "pl_hi", + "run_slug": "h_mid__d_lo__pl_hi", + "test_loss": 1.9772688150405884, + "test_best_rerun_accuracy": 0.3893727660179138, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.3001757264137268, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.2939109206199646, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.33165252208709717, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.33337152004241943, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.399419367313385, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.3965543508529663, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.3988463580608368, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.5517610311508179, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.566315233707428, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.6116204261779785, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.6505462527275085, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s42_community_detection__community_detection__05__h_mid__d_lo__pl_hi__s42" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0.4-0.6__deg_1-2.5__gamma_4-5__s43", + "train_seed": 43, + "homophily": "h_mid", + "avg_degree": "d_lo", + "power_law": "pl_hi", + "run_slug": "h_mid__d_lo__pl_hi", + "test_loss": 1.993801474571228, + "test_best_rerun_accuracy": 0.38792115449905396, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.30025210976600647, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.29360532760620117, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.3337535262107849, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.3354725241661072, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.39922836422920227, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.3977767527103424, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.4036213755607605, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.5467950105667114, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.5651692152023315, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.6097104549407959, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.6521888375282288, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s43_community_detection__community_detection__05__h_mid__d_lo__pl_hi__s43" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0.4-0.6__deg_1-2.5__gamma_4-5__s44", + "train_seed": 44, + "homophily": "h_mid", + "avg_degree": "d_lo", + "power_law": "pl_hi", + "run_slug": "h_mid__d_lo__pl_hi", + "test_loss": 1.9878644943237305, + "test_best_rerun_accuracy": 0.39200854301452637, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.302582323551178, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.2940255105495453, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.33191993832588196, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.33489954471588135, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.39972496032714844, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.3977385461330414, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.4018641710281372, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.5473679900169373, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.5636793971061707, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.6095576286315918, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.6524562835693359, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s44_community_detection__community_detection__05__h_mid__d_lo__pl_hi__s44" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0.4-0.6__deg_4-5__gamma_1.5-2__s42", + "train_seed": 42, + "homophily": "h_mid", + "avg_degree": "d_hi", + "power_law": "pl_lo", + "run_slug": "h_mid__d_hi__pl_lo", + "test_loss": 1.8586361408233643, + "test_best_rerun_accuracy": 0.4339521825313568, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.2720223069190979, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.2612498998641968, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.31247612833976746, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.3133929371833801, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.3969745635986328, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.3794025480747223, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.4368171691894531, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.5506914258003235, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.5674230456352234, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.6389716267585754, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.6830926537513733, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s42_community_detection__community_detection__06__h_mid__d_hi__pl_lo__s42" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0.4-0.6__deg_4-5__gamma_1.5-2__s43", + "train_seed": 43, + "homophily": "h_mid", + "avg_degree": "d_hi", + "power_law": "pl_lo", + "run_slug": "h_mid__d_hi__pl_lo", + "test_loss": 1.858083724975586, + "test_best_rerun_accuracy": 0.4343723654747009, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.27251890301704407, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.2636183202266693, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.31262892484664917, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.3133929371833801, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.3981969654560089, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.3818855583667755, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.43578577041625977, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.5514172315597534, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.5699060559272766, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.6407670378684998, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.6848116517066956, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s43_community_detection__community_detection__06__h_mid__d_hi__pl_lo__s43" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0.4-0.6__deg_4-5__gamma_1.5-2__s44", + "train_seed": 44, + "homophily": "h_mid", + "avg_degree": "d_hi", + "power_law": "pl_lo", + "run_slug": "h_mid__d_hi__pl_lo", + "test_loss": 1.8485448360443115, + "test_best_rerun_accuracy": 0.4346015751361847, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.27370309829711914, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.26025670766830444, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.3146153390407562, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.3146153390407562, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.39796775579452515, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.3818473517894745, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.4347161650657654, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.5539002418518066, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.5712048411369324, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.6423714756965637, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.6863396763801575, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s44_community_detection__community_detection__06__h_mid__d_hi__pl_lo__s44" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0.4-0.6__deg_4-5__gamma_4-5__s42", + "train_seed": 42, + "homophily": "h_mid", + "avg_degree": "d_hi", + "power_law": "pl_hi", + "run_slug": "h_mid__d_hi__pl_hi", + "test_loss": 1.8221337795257568, + "test_best_rerun_accuracy": 0.4431583881378174, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.2592635154724121, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.2465810924768448, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.2977309226989746, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.2982275187969208, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.38666054606437683, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.3724883496761322, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.4268851578235626, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.5498892068862915, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.5669264197349548, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.6368706822395325, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.6908854842185974, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s42_community_detection__community_detection__07__h_mid__d_hi__pl_hi__s42" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0.4-0.6__deg_4-5__gamma_4-5__s43", + "train_seed": 43, + "homophily": "h_mid", + "avg_degree": "d_hi", + "power_law": "pl_hi", + "run_slug": "h_mid__d_hi__pl_hi", + "test_loss": 1.8122713565826416, + "test_best_rerun_accuracy": 0.4466727674007416, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.258002907037735, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.24883489310741425, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.3008633255958557, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.2982275187969208, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.38899075984954834, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.3786385655403137, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.43089616298675537, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.5534800291061401, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.5708610415458679, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.6460004448890686, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.6990602612495422, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s43_community_detection__community_detection__07__h_mid__d_hi__pl_hi__s43" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0.4-0.6__deg_4-5__gamma_4-5__s44", + "train_seed": 44, + "homophily": "h_mid", + "avg_degree": "d_hi", + "power_law": "pl_hi", + "run_slug": "h_mid__d_hi__pl_hi", + "test_loss": 1.8095487356185913, + "test_best_rerun_accuracy": 0.4467109739780426, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.26105889678001404, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.24956069886684418, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.3012453317642212, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.3008251190185547, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.3903277516365051, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.3787531554698944, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.43357017636299133, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.5560775995254517, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.5717778205871582, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.6448162794113159, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.6958515048027039, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s44_community_detection__community_detection__07__h_mid__d_hi__pl_hi__s44" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0.9-1__deg_1-2.5__gamma_1.5-2__s42", + "train_seed": 42, + "homophily": "h_hi", + "avg_degree": "d_lo", + "power_law": "pl_lo", + "run_slug": "h_hi__d_lo__pl_lo", + "test_loss": 1.3534891605377197, + "test_best_rerun_accuracy": 0.6125754714012146, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.2691954970359802, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.26014211773872375, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.3091527223587036, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.3112919330596924, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.3944151699542999, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.3784857392311096, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.37294673919677734, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.368171751499176, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.6270150542259216, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.6884024739265442, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.7243868708610535, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s42_community_detection__community_detection__08__h_hi__d_lo__pl_lo__s42" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0.9-1__deg_1-2.5__gamma_1.5-2__s43", + "train_seed": 43, + "homophily": "h_hi", + "avg_degree": "d_lo", + "power_law": "pl_lo", + "run_slug": "h_hi__d_lo__pl_lo", + "test_loss": 1.3625643253326416, + "test_best_rerun_accuracy": 0.6104744672775269, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.2709527015686035, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.2596836984157562, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.310184121131897, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.3111773133277893, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.39292535185813904, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.3769959509372711, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.3753151595592499, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.37294673919677734, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.6266330480575562, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.686721682548523, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.724004864692688, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s43_community_detection__community_detection__08__h_hi__d_lo__pl_lo__s43" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0.9-1__deg_1-2.5__gamma_1.5-2__s44", + "train_seed": 44, + "homophily": "h_hi", + "avg_degree": "d_lo", + "power_law": "pl_lo", + "run_slug": "h_hi__d_lo__pl_lo", + "test_loss": 1.3738481998443604, + "test_best_rerun_accuracy": 0.6108946204185486, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.2619374990463257, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.25357168912887573, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.31029871106147766, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.3108717203140259, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.39090076088905334, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.37623193860054016, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.3755825459957123, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.3733287453651428, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.6251814365386963, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.6858812570571899, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.7237374782562256, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s44_community_detection__community_detection__08__h_hi__d_lo__pl_lo__s44" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0.9-1__deg_1-2.5__gamma_4-5__s42", + "train_seed": 42, + "homophily": "h_hi", + "avg_degree": "d_lo", + "power_law": "pl_hi", + "run_slug": "h_hi__d_lo__pl_hi", + "test_loss": 1.2887213230133057, + "test_best_rerun_accuracy": 0.631790041923523, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.2641149163246155, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.25521430373191833, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.3088853359222412, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.31236153841018677, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.39414775371551514, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.3794025480747223, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.3743601441383362, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.3718007504940033, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.6168538331985474, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.6918022632598877, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.728703498840332, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s42_community_detection__community_detection__09__h_hi__d_lo__pl_hi__s42" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0.9-1__deg_1-2.5__gamma_4-5__s43", + "train_seed": 43, + "homophily": "h_hi", + "avg_degree": "d_lo", + "power_law": "pl_hi", + "run_slug": "h_hi__d_lo__pl_hi", + "test_loss": 1.3131598234176636, + "test_best_rerun_accuracy": 0.6270914673805237, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.26273971796035767, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.25043928623199463, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.30430132150650024, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.30854153633117676, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.3867751657962799, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.37393996119499207, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.37054014205932617, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.37485674023628235, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.6096722483634949, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.6884024739265442, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.7274047136306763, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s43_community_detection__community_detection__09__h_hi__d_lo__pl_hi__s43" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0.9-1__deg_1-2.5__gamma_4-5__s44", + "train_seed": 44, + "homophily": "h_hi", + "avg_degree": "d_lo", + "power_law": "pl_hi", + "run_slug": "h_hi__d_lo__pl_hi", + "test_loss": 1.2969404458999634, + "test_best_rerun_accuracy": 0.631866455078125, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.2638475000858307, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.2528840899467468, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.3087325096130371, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.30582931637763977, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.3931163549423218, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.37963175773620605, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.37485674023628235, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.37256476283073425, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.6139506697654724, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.6902360916137695, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.728703498840332, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s44_community_detection__community_detection__09__h_hi__d_lo__pl_hi__s44" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0.9-1__deg_4-5__gamma_1.5-2__s42", + "train_seed": 42, + "homophily": "h_hi", + "avg_degree": "d_hi", + "power_law": "pl_lo", + "run_slug": "h_hi__d_hi__pl_lo", + "test_loss": 1.0757602453231812, + "test_best_rerun_accuracy": 0.704217255115509, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.2512415051460266, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.23997250199317932, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.2966613173484802, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.2991825342178345, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.39685994386672974, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.3777981400489807, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.40041255950927734, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.39391854405403137, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.6141416430473328, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.6303766369819641, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.7415004968643188, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s42_community_detection__community_detection__10__h_hi__d_hi__pl_lo__s42" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0.9-1__deg_4-5__gamma_1.5-2__s43", + "train_seed": 43, + "homophily": "h_hi", + "avg_degree": "d_hi", + "power_law": "pl_lo", + "run_slug": "h_hi__d_hi__pl_lo", + "test_loss": 1.080308198928833, + "test_best_rerun_accuracy": 0.7037206888198853, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.2544884979724884, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.24077469110488892, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.2994499206542969, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.2993735074996948, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.39483535289764404, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.3779127597808838, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.39880815148353577, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.3957521617412567, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.6109710335731506, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.6293834447860718, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.7383680939674377, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s43_community_detection__community_detection__10__h_hi__d_hi__pl_lo__s43" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0.9-1__deg_4-5__gamma_1.5-2__s44", + "train_seed": 44, + "homophily": "h_hi", + "avg_degree": "d_hi", + "power_law": "pl_lo", + "run_slug": "h_hi__d_hi__pl_lo", + "test_loss": 1.0855813026428223, + "test_best_rerun_accuracy": 0.7029566764831543, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.2512032985687256, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.23794789612293243, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.29532432556152344, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.2945603132247925, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.39418596029281616, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.37691953778266907, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.4022461473941803, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.3946443498134613, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.6097486615180969, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.6269004344940186, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.740927517414093, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s44_community_detection__community_detection__10__h_hi__d_hi__pl_lo__s44" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0.9-1__deg_4-5__gamma_4-5__s42", + "train_seed": 42, + "homophily": "h_hi", + "avg_degree": "d_hi", + "power_law": "pl_hi", + "run_slug": "h_hi__d_hi__pl_hi", + "test_loss": 0.9378951191902161, + "test_best_rerun_accuracy": 0.7414240837097168, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.23565588891506195, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.22495989501476288, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.2869585156440735, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.28501030802726746, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.3824203610420227, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.3650393486022949, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.39823517203330994, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.40778517723083496, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.6010008454322815, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.6173886656761169, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.694438099861145, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s42_community_detection__community_detection__11__h_hi__d_hi__pl_hi__s42" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0.9-1__deg_4-5__gamma_4-5__s43", + "train_seed": 43, + "homophily": "h_hi", + "avg_degree": "d_hi", + "power_law": "pl_hi", + "run_slug": "h_hi__d_hi__pl_hi", + "test_loss": 0.9291359782218933, + "test_best_rerun_accuracy": 0.7421116828918457, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.23810069262981415, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.22690808773040771, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.28256550431251526, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.28527772426605225, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.38601115345954895, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.3704637587070465, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.39361295104026794, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.4067537486553192, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.6050118207931519, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.6179998517036438, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.6978760957717896, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s43_community_detection__community_detection__11__h_hi__d_hi__pl_hi__s43" + }, + { + "experiment": "community_detection", + "wandb_project": "shard_cd", + "wandb_run_name": "gread_hom_0.9-1__deg_4-5__gamma_4-5__s44", + "train_seed": 44, + "homophily": "h_hi", + "avg_degree": "d_hi", + "power_law": "pl_hi", + "run_slug": "h_hi__d_hi__pl_hi", + "test_loss": 0.9276243448257446, + "test_best_rerun_accuracy": 0.7423790693283081, + "test_best_rerun_mse": null, + "test_triangles_total_structural": null, + "test_mse_by_total_triangles": null, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.24023990333080292, + "test_best_rerun_mse": null + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.22912369668483734, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.2887539267539978, + "test_best_rerun_mse": null + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.2856215238571167, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.3883795440196991, + "test_best_rerun_mse": null + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.37241193652153015, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.3981969654560089, + "test_best_rerun_mse": null + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": 0.4085109531879425, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": 0.6047062277793884, + "test_best_rerun_mse": null + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": 0.6205974221229553, + "test_best_rerun_mse": null + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": 0.698296308517456, + "test_best_rerun_mse": null + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s44_community_detection__community_detection__11__h_hi__d_hi__pl_hi__s44" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0-0.1__deg_1-2.5__gamma_1.5-2__s42", + "train_seed": 42, + "homophily": "h_lo", + "avg_degree": "d_lo", + "power_law": "pl_lo", + "run_slug": "h_lo__d_lo__pl_lo", + "test_loss": 103.7693099975586, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 99.37831115722656, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 0.07159820688560992, + "ood_test": { + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 27.330459594726562, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 0.14384452418277138 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 10635.9609375, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 0.8066712883959044 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 354.98895263671875, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 0.11964575417482937 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 7336.1064453125, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 0.9801077415247161 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 153.3765411376953, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 0.13244951739006502 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 170894.453125, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 3.968383176783392 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 15196.38671875, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 1.0655158265846305 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 46248.25390625, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 2.3706112002793582 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 3671.532470703125, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 0.6527168836805556 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 756233.25, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 8.5543844665905 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 262601.53125, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 4.369918813339324 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s42_triangle_counting__triangle_counting__00__h_lo__d_lo__pl_lo__s42" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0-0.1__deg_1-2.5__gamma_1.5-2__s43", + "train_seed": 43, + "homophily": "h_lo", + "avg_degree": "d_lo", + "power_law": "pl_lo", + "run_slug": "h_lo__d_lo__pl_lo", + "test_loss": 66.18878936767578, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 65.18682098388672, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 0.04696456843219504, + "ood_test": { + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 27.961711883544922, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 0.14716690465023644 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 8967.6962890625, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 0.6801438216960561 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 183.81419372558594, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 0.06195287958395212 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 7049.56494140625, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 0.9418256434744489 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 153.53924560546875, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 0.13259002211180376 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 165857.828125, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 3.8514264379760355 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 14364.3720703125, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 1.007177960336033 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 46258.82421875, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 2.371153017517556 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 3738.55419921875, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 0.6646318576388889 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 753396.875, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 8.522299865389183 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 261510.90625, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 4.351769860882299 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s43_triangle_counting__triangle_counting__00__h_lo__d_lo__pl_lo__s43" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0-0.1__deg_1-2.5__gamma_1.5-2__s44", + "train_seed": 44, + "homophily": "h_lo", + "avg_degree": "d_lo", + "power_law": "pl_lo", + "run_slug": "h_lo__d_lo__pl_lo", + "test_loss": 87.03919219970703, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 89.05577850341797, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 0.06416122370563254, + "ood_test": { + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 28.453157424926758, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 0.14975346013119345 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 9831.416015625, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 0.7456515749431172 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 255.51260375976562, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 0.08611816776534062 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 7185.37060546875, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 0.9599693527680361 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 167.41357421875, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 0.14457130761550085 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 168127.921875, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 3.904140857212521 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 14739.2236328125, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 1.0334611998886902 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 46426.19140625, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 2.379731990683787 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 3757.74462890625, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 0.6680434895833334 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 754443.6875, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 8.534141233894777 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 261495.125, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 4.3515072471003275 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s44_triangle_counting__triangle_counting__00__h_lo__d_lo__pl_lo__s44" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0-0.1__deg_1-2.5__gamma_4-5__s42", + "train_seed": 42, + "homophily": "h_lo", + "avg_degree": "d_lo", + "power_law": "pl_hi", + "run_slug": "h_lo__d_lo__pl_hi", + "test_loss": 1.9937427043914795, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 2.0566940307617188, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 0.010824705425061677, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 207.88951110839844, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 0.14977630483314008 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 12903.9619140625, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 0.9786850143392112 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 345.58203125, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 0.11647523803505225 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 8225.1513671875, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 1.0988846181947227 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 178.46937561035156, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 0.15411863178786836 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 178153.328125, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 4.136943343047557 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 15967.8134765625, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 1.1196054884702356 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 47949.5390625, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 2.457816344379517 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 4009.41455078125, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 0.7127848090277777 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 765415.875, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 8.658256789927943 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 266707.03125, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 4.438237918725974 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s42_triangle_counting__triangle_counting__01__h_lo__d_lo__pl_hi__s42" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0-0.1__deg_1-2.5__gamma_4-5__s43", + "train_seed": 43, + "homophily": "h_lo", + "avg_degree": "d_lo", + "power_law": "pl_hi", + "run_slug": "h_lo__d_lo__pl_hi", + "test_loss": 3.697826862335205, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 3.676438808441162, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 0.01934967793916401, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 192.7432098388672, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 0.13886398403376599 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 12582.099609375, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 0.9542737663538111 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 313.11309814453125, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 0.10553188343260238 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 8129.0234375, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 1.0860418754175016 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 177.4161834716797, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 0.1532091394401379 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 177003.765625, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 4.110249062441947 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 15768.5087890625, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 1.1056309626323446 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 47807.42578125, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 2.450531845878825 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 4005.360107421875, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 0.7120640190972222 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 764026.0, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 8.64253475560784 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 266085.96875, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 4.427902896344 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s43_triangle_counting__triangle_counting__01__h_lo__d_lo__pl_hi__s43" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0-0.1__deg_1-2.5__gamma_4-5__s44", + "train_seed": 44, + "homophily": "h_lo", + "avg_degree": "d_lo", + "power_law": "pl_hi", + "run_slug": "h_lo__d_lo__pl_hi", + "test_loss": 2.7298684120178223, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 2.851917266845703, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 0.01501009087813528, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 196.38729858398438, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 0.14148940820171785 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 12713.6474609375, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 0.9642508502796738 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 330.8277282714844, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 0.11150243622227313 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 8166.9814453125, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 1.0911130855460922 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 178.12557983398438, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 0.15382174424350983 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 177430.03125, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 4.120147483977336 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 15846.5185546875, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 1.1111007260333403 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 47947.03125, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 2.4576877979394127 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 4023.205078125, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 0.7152364583333334 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 765036.125, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 8.653961121228917 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 266491.375, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 4.434649210390561 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s44_triangle_counting__triangle_counting__01__h_lo__d_lo__pl_hi__s44" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0-0.1__deg_4-5__gamma_1.5-2__s42", + "train_seed": 42, + "homophily": "h_lo", + "avg_degree": "d_hi", + "power_law": "pl_lo", + "run_slug": "h_lo__d_hi__pl_lo", + "test_loss": 3718.912841796875, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 3540.4287109375, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 0.2685194320013273, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 6885.89794921875, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 4.961021577246938 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 4862.27392578125, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 25.590915398848683 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 1078.793701171875, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 0.36359747258910513 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 7947.81689453125, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 1.0618325844397127 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 5710.18359375, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 4.931073915155441 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 116063.1328125, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 2.6951312653840795 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 7656.59765625, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 0.5368530119373159 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 28981.654296875, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 1.4855530420254754 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 5319.77734375, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 0.9457381944444444 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 626065.375, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 7.081947162426615 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 194029.90625, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 3.2288270888456228 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s42_triangle_counting__triangle_counting__02__h_lo__d_hi__pl_lo__s42" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0-0.1__deg_4-5__gamma_1.5-2__s43", + "train_seed": 43, + "homophily": "h_lo", + "avg_degree": "d_hi", + "power_law": "pl_lo", + "run_slug": "h_lo__d_hi__pl_lo", + "test_loss": 2281.53662109375, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 2323.655517578125, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 0.1762347756979996, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 1048.82421875, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 0.7556370452089337 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 1730.8511962890625, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 9.109743138363488 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 1190.6556396484375, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 0.4012995078019675 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 4363.033203125, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 0.5829035675517702 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 1287.7236328125, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 1.1120238625323835 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 108183.5625, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 2.512157776797325 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 10481.5517578125, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 0.7349286045303954 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 44005.53515625, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 2.255653039943103 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 5179.33203125, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 0.9207701388888889 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 713468.625, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 8.070638157076118 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 250656.609375, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 4.171144881683391 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s43_triangle_counting__triangle_counting__02__h_lo__d_hi__pl_lo__s43" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0-0.1__deg_4-5__gamma_1.5-2__s44", + "train_seed": 44, + "homophily": "h_lo", + "avg_degree": "d_hi", + "power_law": "pl_lo", + "run_slug": "h_lo__d_hi__pl_lo", + "test_loss": 3504.039794921875, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 3353.7314453125, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 0.2543596090491088, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 7857.9091796875, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 5.661317852800792 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 5539.9296875, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 29.15752467105263 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 840.58740234375, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 0.2833122353703236 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 7941.345703125, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 1.0609680298096191 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 6454.15771484375, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 5.573538613854707 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 114943.015625, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 2.669120741802898 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 7433.1689453125, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 0.5211869965862082 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 28244.72265625, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 1.4477791099620687 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 6064.890625, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 1.0782027777777778 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 620222.875, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 7.015857776319808 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 191844.9375, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 3.1924673006839397 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s44_triangle_counting__triangle_counting__02__h_lo__d_hi__pl_lo__s44" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0-0.1__deg_4-5__gamma_4-5__s42", + "train_seed": 42, + "homophily": "h_lo", + "avg_degree": "d_hi", + "power_law": "pl_hi", + "run_slug": "h_lo__d_hi__pl_hi", + "test_loss": 133.80874633789062, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 132.99447631835938, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 0.04482456229132436, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 139.5034942626953, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 0.10050684024689864 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 117.73936462402344, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 0.6196808664422286 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 10639.1103515625, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 0.8069101518060295 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 7254.41162109375, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 0.9691932693512024 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 159.9431915283203, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 0.13812019993810043 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 169733.859375, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 3.941432736740665 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 13573.05859375, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 0.9516939134588417 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 45418.02734375, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 2.328055120393152 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 3447.152099609375, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 0.6128270399305555 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 750142.0, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 8.485481261948124 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 256900.21875, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 4.275043994308821 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s42_triangle_counting__triangle_counting__03__h_lo__d_hi__pl_hi__s42" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0-0.1__deg_4-5__gamma_4-5__s43", + "train_seed": 43, + "homophily": "h_lo", + "avg_degree": "d_hi", + "power_law": "pl_hi", + "run_slug": "h_lo__d_hi__pl_hi", + "test_loss": 145.23092651367188, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 140.07705688476562, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 0.04721168078354082, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 126.6436996459961, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 0.09124185853457932 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 89.44583129882812, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 0.470767533151727 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 10464.98046875, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 0.7937034864429275 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 7308.833984375, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 0.9764641261690047 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 153.86282348632812, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 0.1328694503336167 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 169703.109375, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 3.9407186832389 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 13670.0849609375, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 0.9584970523725634 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 45648.53125, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 2.3398703803372802 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 3530.629150390625, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 0.6276674045138889 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 751388.0625, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 8.4995765132405 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 257795.140625, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 4.289936275855757 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s43_triangle_counting__triangle_counting__03__h_lo__d_hi__pl_hi__s43" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0-0.1__deg_4-5__gamma_4-5__s44", + "train_seed": 44, + "homophily": "h_lo", + "avg_degree": "d_hi", + "power_law": "pl_hi", + "run_slug": "h_lo__d_hi__pl_hi", + "test_loss": 145.36387634277344, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 144.36087036132812, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 0.04865550062734349, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 153.95831298828125, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 0.11092097477541876 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 153.75437927246094, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 0.8092335751182155 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 10420.1181640625, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 0.7903009604901403 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 7109.88427734375, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 0.949884338990481 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 186.3960418701172, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 0.1609637667272169 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 169028.328125, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 3.925049417727104 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 13583.720703125, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 0.9524415021122563 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 44931.9921875, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 2.3031417390691478 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 3381.36962890625, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 0.6011323784722222 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 748325.125, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 8.464929074805154 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 256234.21875, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 4.263961172682342 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s44_triangle_counting__triangle_counting__03__h_lo__d_hi__pl_hi__s44" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0.4-0.6__deg_1-2.5__gamma_1.5-2__s42", + "train_seed": 42, + "homophily": "h_mid", + "avg_degree": "d_lo", + "power_law": "pl_lo", + "run_slug": "h_mid__d_lo__pl_lo", + "test_loss": 5268.91015625, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 5439.87353515625, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 0.7267700113769205, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 1550.5223388671875, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 1.1170910222386077 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 2207.710693359375, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 11.619529965049342 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 6727.1220703125, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 0.5102102442406143 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 1256.0379638671875, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 0.4233360174813574 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 1797.516845703125, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 1.5522597976710923 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 150585.78125, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 3.4967903875626973 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 9768.8115234375, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 0.6849538299984224 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 36949.6328125, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 1.89397882067251 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 2867.781005859375, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 0.509827734375 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 706071.125, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 7.986958870174089 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 231655.515625, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 3.854950087780607 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s42_triangle_counting__triangle_counting__04__h_mid__d_lo__pl_lo__s42" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0.4-0.6__deg_1-2.5__gamma_1.5-2__s43", + "train_seed": 43, + "homophily": "h_mid", + "avg_degree": "d_lo", + "power_law": "pl_lo", + "run_slug": "h_mid__d_lo__pl_lo", + "test_loss": 4706.82421875, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 4811.73388671875, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 0.6428502186664997, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 1886.958740234375, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 1.3594803603994057 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 2206.589111328125, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 11.613626901726974 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 4386.828125, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 0.33271354759196053 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 9888.755859375, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 3.3329140072042467 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 1536.0865478515625, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 1.3264996095436636 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 127173.3203125, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 2.953123730087776 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 8226.447265625, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 0.5768088112203759 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 38221.07421875, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 1.9591508646650264 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 2612.12646484375, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 0.46437803819444445 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 692534.75, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 7.833837652568352 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 217828.421875, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 3.6248551723994473 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s43_triangle_counting__triangle_counting__04__h_mid__d_lo__pl_lo__s43" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0.4-0.6__deg_1-2.5__gamma_1.5-2__s44", + "train_seed": 44, + "homophily": "h_mid", + "avg_degree": "d_lo", + "power_law": "pl_lo", + "run_slug": "h_mid__d_lo__pl_lo", + "test_loss": 5240.5751953125, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 5442.1123046875, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 0.7270691121826988, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 1381.1591796875, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 0.995071455106268 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 1802.9298095703125, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 9.489104260896381 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 6106.7958984375, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 0.46316237379124003 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 1820.029541015625, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 0.613424179648003 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 1440.3955078125, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 1.24386485994171 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 147431.1875, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 3.4235367708526843 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 9446.3076171875, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 0.662341019295155 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 38090.96875, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 1.9524818673432776 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 2799.69140625, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 0.4977229166666667 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 707662.5, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 8.00496023890592 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 232313.203125, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 3.865894582147671 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s44_triangle_counting__triangle_counting__04__h_mid__d_lo__pl_lo__s44" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0.4-0.6__deg_1-2.5__gamma_4-5__s42", + "train_seed": 42, + "homophily": "h_mid", + "avg_degree": "d_lo", + "power_law": "pl_hi", + "run_slug": "h_mid__d_lo__pl_hi", + "test_loss": 131.87379455566406, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 139.33749389648438, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 0.12032598782079826, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 142.73898315429688, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 0.10283788411692858 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 40.2872200012207, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 0.21203800000642475 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 11467.9228515625, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 0.8697704096748199 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 152.51419067382812, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 0.05140350208083186 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 7618.44482421875, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 1.0178282998288244 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 172906.84375, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 4.015113406789894 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 14353.3798828125, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 1.0064072277950147 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 46522.81640625, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 2.38468483296171 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 3646.09130859375, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 0.6481940104166667 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 756737.9375, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 8.560093407463548 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 260920.59375, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 4.3419465453547 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s42_triangle_counting__triangle_counting__05__h_mid__d_lo__pl_hi__s42" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0.4-0.6__deg_1-2.5__gamma_4-5__s43", + "train_seed": 43, + "homophily": "h_mid", + "avg_degree": "d_lo", + "power_law": "pl_hi", + "run_slug": "h_mid__d_lo__pl_hi", + "test_loss": 139.33627319335938, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 144.46835327148438, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 0.1247567817543043, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 140.94012451171875, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 0.10154187644936509 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 46.47947692871094, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 0.24462882594058388 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 11487.3427734375, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 0.8712432896046644 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 178.04525756835938, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 0.06000851283058961 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 7571.34912109375, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 1.01153628872328 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 173010.296875, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 4.017515717885009 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 14542.212890625, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 1.0196475172223392 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 46360.8046875, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 2.376380372520375 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 3662.4609375, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 0.6511041666666667 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 756791.8125, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 8.560702832483061 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 261337.4375, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 4.348883189389779 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s43_triangle_counting__triangle_counting__05__h_mid__d_lo__pl_hi__s43" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0.4-0.6__deg_1-2.5__gamma_4-5__s44", + "train_seed": 44, + "homophily": "h_mid", + "avg_degree": "d_lo", + "power_law": "pl_hi", + "run_slug": "h_mid__d_lo__pl_hi", + "test_loss": 143.45443725585938, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 150.46644592285156, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 0.12993648179866282, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 147.7235870361328, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 0.10642909728828012 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 35.235042572021484, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 0.1854475924843236 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 11706.271484375, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 0.8878476666192643 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 196.869384765625, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 0.06635301138039265 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 7680.59423828125, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 1.0261314947603541 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 173806.390625, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 4.036002011540963 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 14801.11328125, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 1.0378006788143317 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 46629.6328125, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 2.3901600703521453 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 3709.257080078125, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 0.6594234809027778 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 758243.3125, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 8.57712195853082 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 262258.90625, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 4.364217234120447 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s44_triangle_counting__triangle_counting__05__h_mid__d_lo__pl_hi__s44" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0.4-0.6__deg_4-5__gamma_1.5-2__s42", + "train_seed": 42, + "homophily": "h_mid", + "avg_degree": "d_hi", + "power_law": "pl_lo", + "run_slug": "h_mid__d_hi__pl_lo", + "test_loss": 112329.3515625, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 78939.2265625, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 1.8330676797905443, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 73191.6171875, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 52.73171267110951 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 80139.953125, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 421.7892269736842 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 38812.30859375, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 2.943671489855897 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 71941.7734375, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 24.24731157313785 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 56479.47265625, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 7.545687729625919 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 76361.3828125, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 65.94247220423144 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 45758.6640625, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 3.2084324822956107 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 46285.53125, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 2.37252197703624 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 62872.69921875, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 11.17736875 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 472597.21875, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 5.345940960714003 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 122115.7890625, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 2.0321133753099363 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s42_triangle_counting__triangle_counting__06__h_mid__d_hi__pl_lo__s42" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0.4-0.6__deg_4-5__gamma_1.5-2__s43", + "train_seed": 43, + "homophily": "h_mid", + "avg_degree": "d_hi", + "power_law": "pl_lo", + "run_slug": "h_mid__d_hi__pl_lo", + "test_loss": 113411.0859375, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 79427.1328125, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 1.8443974738180382, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 67665.515625, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 48.75037148775216 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 73816.90625, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 388.51003289473687 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 35123.19921875, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 2.6638755569776262 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 65356.1328125, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 22.027682107347488 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 51817.13671875, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 6.922797156813627 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 70380.8671875, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 60.77795093911917 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 41154.4765625, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 2.88560346112046 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 43288.67578125, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 2.218907979970783 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 57791.828125, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 10.274102777777777 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 480181.34375, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 5.431731318507291 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 124360.84375, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 2.0694730459454513 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s43_triangle_counting__triangle_counting__06__h_mid__d_hi__pl_lo__s43" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0.4-0.6__deg_4-5__gamma_1.5-2__s44", + "train_seed": 44, + "homophily": "h_mid", + "avg_degree": "d_hi", + "power_law": "pl_lo", + "run_slug": "h_mid__d_hi__pl_lo", + "test_loss": 112798.171875, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 79224.0703125, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 1.8396821083155304, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 71314.46875, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 51.379300252161386 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 77644.546875, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 408.65550986842106 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 37652.328125, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 2.8556942074326885 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 69081.34375, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 23.283230114593866 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 54969.44140625, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 7.343946747661991 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 74252.359375, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 64.12120844127807 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 43952.16796875, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 3.0817674918489693 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 45301.55859375, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 2.322085119367984 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 61146.87890625, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 10.87055625 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 475524.75, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 5.379056706220378 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 123280.6953125, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 2.051498432637745 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s44_triangle_counting__triangle_counting__06__h_mid__d_hi__pl_lo__s44" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0.4-0.6__deg_4-5__gamma_4-5__s42", + "train_seed": 42, + "homophily": "h_mid", + "avg_degree": "d_hi", + "power_law": "pl_hi", + "run_slug": "h_mid__d_hi__pl_hi", + "test_loss": 7600.79150390625, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 7335.857421875, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 0.5143638635447343, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 6417.23779296875, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 4.623370167844921 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 6146.603515625, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 32.35054481907895 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 5137.0224609375, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 0.38961110814846417 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 2307.953125, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 0.7778743259184361 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 7319.19677734375, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 0.9778486008475283 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 6568.56494140625, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 5.672335873407815 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 119813.5078125, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 2.782219668690786 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 28872.88671875, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 1.4799777906991645 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 5654.92333984375, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 1.005319704861111 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 622963.0625, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 7.046854320554733 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 188370.5, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 3.134649626412394 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s42_triangle_counting__triangle_counting__07__h_mid__d_hi__pl_hi__s42" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0.4-0.6__deg_4-5__gamma_4-5__s43", + "train_seed": 43, + "homophily": "h_mid", + "avg_degree": "d_hi", + "power_law": "pl_hi", + "run_slug": "h_mid__d_hi__pl_hi", + "test_loss": 8644.5048828125, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 8210.6728515625, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 0.5757027662012691, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 4997.88232421875, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 3.6007797724918946 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 6317.9521484375, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 33.25237972861842 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 5031.31298828125, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 0.3815937040789723 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 4128.7373046875, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 1.391552849574486 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 5992.5927734375, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 0.8006135969856379 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 5513.1572265625, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 4.760930247463299 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 134843.890625, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 3.131243976987739 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 31513.509765625, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 1.6153318860846275 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 4650.177734375, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 0.8266982638888889 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 667359.875, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 7.549063662997862 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 210298.40625, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 3.499549136338675 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s43_triangle_counting__triangle_counting__07__h_mid__d_hi__pl_hi__s43" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0.4-0.6__deg_4-5__gamma_4-5__s44", + "train_seed": 44, + "homophily": "h_mid", + "avg_degree": "d_hi", + "power_law": "pl_hi", + "run_slug": "h_mid__d_hi__pl_hi", + "test_loss": 8358.1484375, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 8036.90576171875, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 0.5635188446023524, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 5527.56103515625, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 3.9823926766255404 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 6695.72265625, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 35.24064555921053 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 5281.51171875, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 0.4005697170079636 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 4290.89404296875, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 1.4462062834407652 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 6490.30126953125, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 0.8671077180402471 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 5944.7890625, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 5.133669311312608 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 133170.0, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 3.0923741408136727 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 31789.173828125, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 1.6294619830911374 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 4989.509765625, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 0.8870239583333334 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 663439.75, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 7.504719862448107 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 207949.40625, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 3.460459724926364 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s44_triangle_counting__triangle_counting__07__h_mid__d_hi__pl_hi__s44" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0.9-1__deg_1-2.5__gamma_1.5-2__s42", + "train_seed": 42, + "homophily": "h_hi", + "avg_degree": "d_lo", + "power_law": "pl_lo", + "run_slug": "h_hi__d_lo__pl_lo", + "test_loss": 27631.06640625, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 28041.640625, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 1.4373694512788968, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 12090.3095703125, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 8.710597673135807 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 14401.25, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 75.79605263157895 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 5846.75048828125, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 0.443439551633011 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 13098.5966796875, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 4.414761267168015 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 9427.224609375, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 1.259482245741483 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 12788.3828125, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 11.043508473661486 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 115433.5859375, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 2.680512398697288 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 9602.6943359375, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 0.673306291960279 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 9366.4970703125, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 1.6651550347222221 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 619468.25, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 7.007321584109136 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 183960.1875, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 3.061258174829015 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s42_triangle_counting__triangle_counting__08__h_hi__d_lo__pl_lo__s42" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0.9-1__deg_1-2.5__gamma_1.5-2__s43", + "train_seed": 43, + "homophily": "h_hi", + "avg_degree": "d_lo", + "power_law": "pl_lo", + "run_slug": "h_hi__d_lo__pl_lo", + "test_loss": 26880.51171875, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 27414.375, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 1.405216823004767, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 14308.962890625, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 10.30905107393732 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 16664.80859375, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 87.70951891447369 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 7211.45068359375, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 0.5469435482437429 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 17475.765625, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 5.89004571115605 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 10412.0283203125, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 1.3910525478039413 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 14742.48828125, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 12.730991607297064 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 109808.765625, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 2.549897028260264 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 10939.78515625, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 0.7670582776784463 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 10624.6435546875, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 1.8888255208333333 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 607455.4375, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 6.871434651538975 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 176417.859375, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 2.9357472480155757 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s43_triangle_counting__triangle_counting__08__h_hi__d_lo__pl_lo__s43" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0.9-1__deg_1-2.5__gamma_1.5-2__s44", + "train_seed": 44, + "homophily": "h_hi", + "avg_degree": "d_lo", + "power_law": "pl_lo", + "run_slug": "h_hi__d_lo__pl_lo", + "test_loss": 27448.056640625, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 27911.796875, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 1.430713869239838, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 12597.525390625, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 9.076026938490633 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 14665.5302734375, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 77.18700143914474 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 6016.37548828125, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 0.456304549736917 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 13118.1416015625, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 4.421348702919616 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 9753.3037109375, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 1.3030465879676019 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 13235.19140625, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 11.429353545984457 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 114429.4140625, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 2.6571942704463125 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 9716.5771484375, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 0.681291344021701 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 9799.826171875, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 1.7421913194444445 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 617749.9375, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 6.9878843195366676 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 183773.703125, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 3.05815491196978 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s44_triangle_counting__triangle_counting__08__h_hi__d_lo__pl_lo__s44" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0.9-1__deg_1-2.5__gamma_4-5__s42", + "train_seed": 42, + "homophily": "h_hi", + "avg_degree": "d_lo", + "power_law": "pl_hi", + "run_slug": "h_hi__d_lo__pl_hi", + "test_loss": 2523.91259765625, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 2642.5126953125, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 0.46978003472222224, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 862.0759887695312, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 0.6210922109290571 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 1253.0223388671875, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 6.59485441509046 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 5847.9228515625, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 0.4435284680745165 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 2602.08203125, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 0.8770077624705089 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 5652.150390625, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 0.7551303127087509 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 858.4324340820312, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 0.7413060743368145 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 147585.4375, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 3.4271186489875536 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 8963.4814453125, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 0.6284869895745688 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 40217.12109375, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 2.0614650209518683 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 711705.4375, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 8.050693273983915 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 231420.0, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 3.8510309021017424 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s42_triangle_counting__triangle_counting__09__h_hi__d_lo__pl_hi__s42" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0.9-1__deg_1-2.5__gamma_4-5__s43", + "train_seed": 43, + "homophily": "h_hi", + "avg_degree": "d_lo", + "power_law": "pl_hi", + "run_slug": "h_hi__d_lo__pl_hi", + "test_loss": 2590.84619140625, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 2738.189208984375, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 0.4867891927083333, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 554.4853515625, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 0.39948512360410665 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 880.0433349609375, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 4.631807026110198 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 7823.33837890625, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 0.5933514128863292 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 641.6730346679688, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 0.21626998135084893 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 5860.53662109375, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 0.7829708244614229 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 651.82861328125, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 0.5628917213136874 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 157703.140625, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 3.6620643838240756 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 10877.1396484375, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 0.7626658006196536 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 40549.3828125, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 2.078496222897124 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 725308.5625, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 8.204569556463015 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 241854.109375, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 4.024663594345431 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s43_triangle_counting__triangle_counting__09__h_hi__d_lo__pl_hi__s43" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0.9-1__deg_1-2.5__gamma_4-5__s44", + "train_seed": 44, + "homophily": "h_hi", + "avg_degree": "d_lo", + "power_law": "pl_hi", + "run_slug": "h_hi__d_lo__pl_hi", + "test_loss": 2676.309326171875, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 2828.345947265625, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 0.5028170572916667, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 686.5233154296875, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 0.4946133396467489 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 969.9566650390625, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 5.105035079152961 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 7182.55908203125, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 0.5447523004953546 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 1123.830810546875, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 0.3787768151489299 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 5947.642578125, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 0.7946082268704074 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 726.1038818359375, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 0.6270327131571135 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 154475.375, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 3.5871116245587964 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 10376.126953125, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 0.7275365974705511 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 40826.7890625, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 2.092715621636168 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 722347.4375, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 8.171073804056423 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 239842.140625, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 3.991182677266903 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s44_triangle_counting__triangle_counting__09__h_hi__d_lo__pl_hi__s44" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0.9-1__deg_4-5__gamma_1.5-2__s42", + "train_seed": 42, + "homophily": "h_hi", + "avg_degree": "d_hi", + "power_law": "pl_lo", + "run_slug": "h_hi__d_hi__pl_lo", + "test_loss": 506702.65625, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 325270.21875, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 3.6794024948248363, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 365471.6875, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 263.30813220461096 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 403677.90625, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 2124.620559210526 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 286011.53125, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 21.69219046264695 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 431218.21875, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 145.33812563195147 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 313512.46875, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 41.885433366733466 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 382661.96875, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 330.45075021588946 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 167717.125, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 3.8946016394203977 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 331581.5, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 23.249298836067872 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 245907.59375, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 12.604828220308576 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 338567.375, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 60.18975555555556 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 166656.625, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 2.773311783402393 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s42_triangle_counting__triangle_counting__10__h_hi__d_hi__pl_lo__s42" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0.9-1__deg_4-5__gamma_1.5-2__s43", + "train_seed": 43, + "homophily": "h_hi", + "avg_degree": "d_hi", + "power_law": "pl_lo", + "run_slug": "h_hi__d_hi__pl_lo", + "test_loss": 502976.25, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 323855.53125, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 3.6633997856407587, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 385846.59375, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 277.9874594740634 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 411094.25, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 2163.653947368421 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 281455.28125, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 21.346627322715207 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 390664.375, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 131.6698264239973 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 337288.28125, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 45.06189462257849 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 400468.34375, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 345.82758527633854 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 168898.21875, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 3.922028115130968 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 310284.53125, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 21.756032200953584 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 266071.96875, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 13.638421689989237 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 362176.8125, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 64.3869888888889 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 163764.640625, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 2.7251866377947516 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s43_triangle_counting__triangle_counting__10__h_hi__d_hi__pl_lo__s43" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0.9-1__deg_4-5__gamma_1.5-2__s44", + "train_seed": 44, + "homophily": "h_hi", + "avg_degree": "d_hi", + "power_law": "pl_lo", + "run_slug": "h_hi__d_hi__pl_lo", + "test_loss": 504053.34375, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 323735.75, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 3.662044840107236, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 366068.8125, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 263.73833753602304 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 394403.25, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 2075.8065789473685 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 279357.1875, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 21.1875 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 400534.8125, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 134.99656639703403 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 317624.40625, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 42.434790414161654 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 381594.9375, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 329.5293069948186 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 165287.8125, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 3.8381899614527217 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 313355.6875, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 21.971370600196327 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 248050.796875, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 12.714685369573017 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 340291.71875, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 60.49630555555556 + }, + "h_hi__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 160278.734375, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 2.667178113507397 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s44_triangle_counting__triangle_counting__10__h_hi__d_hi__pl_lo__s44" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0.9-1__deg_4-5__gamma_4-5__s42", + "train_seed": 42, + "homophily": "h_hi", + "avg_degree": "d_hi", + "power_law": "pl_hi", + "run_slug": "h_hi__d_hi__pl_hi", + "test_loss": 87454.625, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 84453.0390625, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 1.4053723239395604, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 75107.2265625, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 54.11183469920749 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 66173.8984375, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 348.2836759868421 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 38652.94140625, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 2.931584482840349 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 29759.92578125, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 10.03030865562858 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 78118.1171875, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 10.43662220273881 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 71284.9765625, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 61.5587016947323 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 72043.921875, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 1.6729500714053502 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 36780.6484375, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 2.5789264084630488 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 67793.890625, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 3.4750059267517557 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 70911.265625, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 12.606447222222222 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 385699.25, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 4.3629656233385745 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s42_triangle_counting__triangle_counting__11__h_hi__d_hi__pl_hi__s42" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0.9-1__deg_4-5__gamma_4-5__s43", + "train_seed": 43, + "homophily": "h_hi", + "avg_degree": "d_hi", + "power_law": "pl_hi", + "run_slug": "h_hi__d_hi__pl_hi", + "test_loss": 96312.2109375, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 84981.21875, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 1.4141616952057643, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 84683.34375, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 61.011054574927954 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 70101.375, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 368.9546052631579 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 32338.546875, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 2.45267704778157 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 24611.18359375, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 8.294972562773845 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 94510.46875, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 12.626649131596526 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 77806.578125, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 67.19048197322971 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 85313.9609375, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 1.9810969937186513 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 27104.462890625, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 1.9004671778589959 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 92073.78125, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 4.719554116049003 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 74638.0859375, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 13.268993055555555 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 390714.125, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 4.419693053403165 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s43_triangle_counting__triangle_counting__11__h_hi__d_hi__pl_hi__s43" + }, + { + "experiment": "triangle_counting", + "wandb_project": "shard_tri", + "wandb_run_name": "gread_hom_0.9-1__deg_4-5__gamma_4-5__s44", + "train_seed": 44, + "homophily": "h_hi", + "avg_degree": "d_hi", + "power_law": "pl_hi", + "run_slug": "h_hi__d_hi__pl_hi", + "test_loss": 77906.46875, + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 79139.1953125, + "test_triangles_total_structural": 60093.0, + "test_mse_by_total_triangles": 1.3169453232905663, + "ood_test": { + "h_lo__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 100867.1953125, + "test_triangles_total_structural": 1388.0, + "test_mse_by_total_triangles": 72.67088999459654 + }, + "h_lo__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 74055.8046875, + "test_triangles_total_structural": 190.0, + "test_mse_by_total_triangles": 389.76739309210524 + }, + "h_lo__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 60992.74609375, + "test_triangles_total_structural": 13185.0, + "test_mse_by_total_triangles": 4.6259193093477435 + }, + "h_lo__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 32375.06640625, + "test_triangles_total_structural": 2967.0, + "test_mse_by_total_triangles": 10.911717696747557 + }, + "h_mid__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 120791.3828125, + "test_triangles_total_structural": 7485.0, + "test_mse_by_total_triangles": 16.13779329492318 + }, + "h_mid__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 98727.5078125, + "test_triangles_total_structural": 1158.0, + "test_mse_by_total_triangles": 85.25691520941278 + }, + "h_mid__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 77641.0859375, + "test_triangles_total_structural": 43064.0, + "test_mse_by_total_triangles": 1.8029232290892625 + }, + "h_mid__d_hi__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 34599.98828125, + "test_triangles_total_structural": 14262.0, + "test_mse_by_total_triangles": 2.4260263834840834 + }, + "h_hi__d_lo__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 94788.75, + "test_triangles_total_structural": 19509.0, + "test_mse_by_total_triangles": 4.858719052744887 + }, + "h_hi__d_lo__pl_hi": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 101391.234375, + "test_triangles_total_structural": 5625.0, + "test_mse_by_total_triangles": 18.025108333333332 + }, + "h_hi__d_hi__pl_lo": { + "test_best_rerun_accuracy": null, + "test_best_rerun_mse": 332930.8125, + "test_triangles_total_structural": 88403.0, + "test_mse_by_total_triangles": 3.766057854371458 + } + }, + "output_dir": "/private/tmp/claude-501/-Users-hina-Documents-topoLab-tdl-challenge-2026/34bdf9a1-f9d0-47d8-984b-c8715484fb64/scratchpad/exp-gread/logs/train/runs/notebook_gu_grid_2026-08-02_gread-vc_s44_triangle_counting__triangle_counting__11__h_hi__d_hi__pl_hi__s44" + } + ] +} diff --git a/2026_tdl_challenge/run_evaluation.ipynb b/2026_tdl_challenge/run_evaluation.ipynb index 8542dbaab..5ab3b385d 100644 --- a/2026_tdl_challenge/run_evaluation.ipynb +++ b/2026_tdl_challenge/run_evaluation.ipynb @@ -104,7 +104,7 @@ "outputs": [], "source": [ "# Your model configuration (e.g., \"graph/gcn\", \"graph/gin\", \"graph/gat\")\n", - "MODEL_CONFIG = \"graph/gin\"" + "MODEL_CONFIG = \"graph/gread\"" ] }, { diff --git a/configs/model/graph/gread.yaml b/configs/model/graph/gread.yaml new file mode 100644 index 000000000..9e0100aae --- /dev/null +++ b/configs/model/graph/gread.yaml @@ -0,0 +1,50 @@ +_target_: topobench.model.TBModel + +model_name: gread +model_domain: graph + +feature_encoder: + _target_: topobench.nn.encoders.${model.feature_encoder.encoder_name} + encoder_name: AllCellFeatureEncoder + in_channels: ${infer_in_channels:${dataset},${oc.select:transforms,null}} + out_channels: 64 + proj_dropout: 0.0 + +backbone: + _target_: topobench.nn.backbones.GREAD + in_channels: ${model.feature_encoder.out_channels} + hidden_channels: ${model.feature_encoder.out_channels} + reaction_term: bspm # GREAD-BS. Options: bspm, fisher, allen-cahn, zeldovich, st, fb, fb3, none + time: 3.0 + step_size: 1.0 + # Per-channel reaction gate (the "(VC)" variants of the paper); every + # tuned dataset configuration in the reference implementation's + # src/gread_params.py sets beta_diag: true. + beta_diag: true + add_source: false + data_norm: rw + self_loop_weight: 1.0 + # Terminal ReLU pinned explicitly for the submitted benchmark runs + # (the constructor default is false, matching the reference CLI default). + xn_activation: true + input_dropout: 0.0 + dropout: 0.0 + +backbone_wrapper: + _target_: topobench.nn.wrappers.GNNWrapper + _partial_: true + wrapper_name: GNNWrapper + out_channels: ${model.feature_encoder.out_channels} + num_cell_dimensions: ${infer_num_cell_dimensions:${oc.select:model.feature_encoder.selected_dimensions,null},${model.feature_encoder.in_channels}} + +readout: + _target_: topobench.nn.readouts.${model.readout.readout_name} + readout_name: NoReadOut # Use in case readout is not needed Options: PropagateSignalDown + num_cell_dimensions: ${infer_num_cell_dimensions:${oc.select:model.feature_encoder.selected_dimensions,null},${model.feature_encoder.in_channels}} # The highest order of cell dimensions to consider + hidden_dim: ${model.feature_encoder.out_channels} + out_channels: ${dataset.parameters.num_classes} + task_level: ${define_task_level:${dataset.parameters.task_level},${dataset.split_params.learning_setting}} # Handles the edge case of node-inductive task + pooling_type: sum + +# compile model for faster training with pytorch 2.0 +compile: false diff --git a/configs/model/graph/gread_allen_cahn.yaml b/configs/model/graph/gread_allen_cahn.yaml new file mode 100644 index 000000000..7ce21327c --- /dev/null +++ b/configs/model/graph/gread_allen_cahn.yaml @@ -0,0 +1,47 @@ +_target_: topobench.model.TBModel + +model_name: gread_allen_cahn +model_domain: graph + +feature_encoder: + _target_: topobench.nn.encoders.${model.feature_encoder.encoder_name} + encoder_name: AllCellFeatureEncoder + in_channels: ${infer_in_channels:${dataset},${oc.select:transforms,null}} + out_channels: 64 + proj_dropout: 0.0 + +backbone: + _target_: topobench.nn.backbones.GREAD + in_channels: ${model.feature_encoder.out_channels} + hidden_channels: ${model.feature_encoder.out_channels} + reaction_term: allen-cahn + time: 3.0 + step_size: 1.0 + beta_diag: false + add_source: false + data_norm: rw + self_loop_weight: 1.0 + # Terminal ReLU pinned explicitly for the submitted benchmark runs + # (the constructor default is false, matching the reference CLI default). + xn_activation: true + input_dropout: 0.0 + dropout: 0.0 + +backbone_wrapper: + _target_: topobench.nn.wrappers.GNNWrapper + _partial_: true + wrapper_name: GNNWrapper + out_channels: ${model.feature_encoder.out_channels} + num_cell_dimensions: ${infer_num_cell_dimensions:${oc.select:model.feature_encoder.selected_dimensions,null},${model.feature_encoder.in_channels}} + +readout: + _target_: topobench.nn.readouts.${model.readout.readout_name} + readout_name: NoReadOut # Use in case readout is not needed Options: PropagateSignalDown + num_cell_dimensions: ${infer_num_cell_dimensions:${oc.select:model.feature_encoder.selected_dimensions,null},${model.feature_encoder.in_channels}} # The highest order of cell dimensions to consider + hidden_dim: ${model.feature_encoder.out_channels} + out_channels: ${dataset.parameters.num_classes} + task_level: ${define_task_level:${dataset.parameters.task_level},${dataset.split_params.learning_setting}} # Handles the edge case of node-inductive task + pooling_type: sum + +# compile model for faster training with pytorch 2.0 +compile: false diff --git a/configs/model/graph/gread_fisher.yaml b/configs/model/graph/gread_fisher.yaml new file mode 100644 index 000000000..1cd3f040d --- /dev/null +++ b/configs/model/graph/gread_fisher.yaml @@ -0,0 +1,47 @@ +_target_: topobench.model.TBModel + +model_name: gread_fisher +model_domain: graph + +feature_encoder: + _target_: topobench.nn.encoders.${model.feature_encoder.encoder_name} + encoder_name: AllCellFeatureEncoder + in_channels: ${infer_in_channels:${dataset},${oc.select:transforms,null}} + out_channels: 64 + proj_dropout: 0.0 + +backbone: + _target_: topobench.nn.backbones.GREAD + in_channels: ${model.feature_encoder.out_channels} + hidden_channels: ${model.feature_encoder.out_channels} + reaction_term: fisher + time: 3.0 + step_size: 1.0 + beta_diag: false + add_source: false + data_norm: rw + self_loop_weight: 1.0 + # Terminal ReLU pinned explicitly for the submitted benchmark runs + # (the constructor default is false, matching the reference CLI default). + xn_activation: true + input_dropout: 0.0 + dropout: 0.0 + +backbone_wrapper: + _target_: topobench.nn.wrappers.GNNWrapper + _partial_: true + wrapper_name: GNNWrapper + out_channels: ${model.feature_encoder.out_channels} + num_cell_dimensions: ${infer_num_cell_dimensions:${oc.select:model.feature_encoder.selected_dimensions,null},${model.feature_encoder.in_channels}} + +readout: + _target_: topobench.nn.readouts.${model.readout.readout_name} + readout_name: NoReadOut # Use in case readout is not needed Options: PropagateSignalDown + num_cell_dimensions: ${infer_num_cell_dimensions:${oc.select:model.feature_encoder.selected_dimensions,null},${model.feature_encoder.in_channels}} # The highest order of cell dimensions to consider + hidden_dim: ${model.feature_encoder.out_channels} + out_channels: ${dataset.parameters.num_classes} + task_level: ${define_task_level:${dataset.parameters.task_level},${dataset.split_params.learning_setting}} # Handles the edge case of node-inductive task + pooling_type: sum + +# compile model for faster training with pytorch 2.0 +compile: false diff --git a/configs/model/graph/gread_source_term.yaml b/configs/model/graph/gread_source_term.yaml new file mode 100644 index 000000000..5afb51593 --- /dev/null +++ b/configs/model/graph/gread_source_term.yaml @@ -0,0 +1,47 @@ +_target_: topobench.model.TBModel + +model_name: gread_source_term +model_domain: graph + +feature_encoder: + _target_: topobench.nn.encoders.${model.feature_encoder.encoder_name} + encoder_name: AllCellFeatureEncoder + in_channels: ${infer_in_channels:${dataset},${oc.select:transforms,null}} + out_channels: 64 + proj_dropout: 0.0 + +backbone: + _target_: topobench.nn.backbones.GREAD + in_channels: ${model.feature_encoder.out_channels} + hidden_channels: ${model.feature_encoder.out_channels} + reaction_term: st + time: 3.0 + step_size: 1.0 + beta_diag: false + add_source: false + data_norm: rw + self_loop_weight: 1.0 + # Terminal ReLU pinned explicitly for the submitted benchmark runs + # (the constructor default is false, matching the reference CLI default). + xn_activation: true + input_dropout: 0.0 + dropout: 0.0 + +backbone_wrapper: + _target_: topobench.nn.wrappers.GNNWrapper + _partial_: true + wrapper_name: GNNWrapper + out_channels: ${model.feature_encoder.out_channels} + num_cell_dimensions: ${infer_num_cell_dimensions:${oc.select:model.feature_encoder.selected_dimensions,null},${model.feature_encoder.in_channels}} + +readout: + _target_: topobench.nn.readouts.${model.readout.readout_name} + readout_name: NoReadOut # Use in case readout is not needed Options: PropagateSignalDown + num_cell_dimensions: ${infer_num_cell_dimensions:${oc.select:model.feature_encoder.selected_dimensions,null},${model.feature_encoder.in_channels}} # The highest order of cell dimensions to consider + hidden_dim: ${model.feature_encoder.out_channels} + out_channels: ${dataset.parameters.num_classes} + task_level: ${define_task_level:${dataset.parameters.task_level},${dataset.split_params.learning_setting}} # Handles the edge case of node-inductive task + pooling_type: sum + +# compile model for faster training with pytorch 2.0 +compile: false diff --git a/configs/model/graph/gread_zeldovich.yaml b/configs/model/graph/gread_zeldovich.yaml new file mode 100644 index 000000000..bdb2f4f7f --- /dev/null +++ b/configs/model/graph/gread_zeldovich.yaml @@ -0,0 +1,47 @@ +_target_: topobench.model.TBModel + +model_name: gread_zeldovich +model_domain: graph + +feature_encoder: + _target_: topobench.nn.encoders.${model.feature_encoder.encoder_name} + encoder_name: AllCellFeatureEncoder + in_channels: ${infer_in_channels:${dataset},${oc.select:transforms,null}} + out_channels: 64 + proj_dropout: 0.0 + +backbone: + _target_: topobench.nn.backbones.GREAD + in_channels: ${model.feature_encoder.out_channels} + hidden_channels: ${model.feature_encoder.out_channels} + reaction_term: zeldovich + time: 3.0 + step_size: 1.0 + beta_diag: false + add_source: false + data_norm: rw + self_loop_weight: 1.0 + # Terminal ReLU pinned explicitly for the submitted benchmark runs + # (the constructor default is false, matching the reference CLI default). + xn_activation: true + input_dropout: 0.0 + dropout: 0.0 + +backbone_wrapper: + _target_: topobench.nn.wrappers.GNNWrapper + _partial_: true + wrapper_name: GNNWrapper + out_channels: ${model.feature_encoder.out_channels} + num_cell_dimensions: ${infer_num_cell_dimensions:${oc.select:model.feature_encoder.selected_dimensions,null},${model.feature_encoder.in_channels}} + +readout: + _target_: topobench.nn.readouts.${model.readout.readout_name} + readout_name: NoReadOut # Use in case readout is not needed Options: PropagateSignalDown + num_cell_dimensions: ${infer_num_cell_dimensions:${oc.select:model.feature_encoder.selected_dimensions,null},${model.feature_encoder.in_channels}} # The highest order of cell dimensions to consider + hidden_dim: ${model.feature_encoder.out_channels} + out_channels: ${dataset.parameters.num_classes} + task_level: ${define_task_level:${dataset.parameters.task_level},${dataset.split_params.learning_setting}} # Handles the edge case of node-inductive task + pooling_type: sum + +# compile model for faster training with pytorch 2.0 +compile: false diff --git a/test/nn/backbones/graph/test_gread.py b/test/nn/backbones/graph/test_gread.py new file mode 100644 index 000000000..cfc4ae8cd --- /dev/null +++ b/test/nn/backbones/graph/test_gread.py @@ -0,0 +1,564 @@ +"""Unit tests for the GREAD backbone.""" + +import pytest +import torch +import torch_geometric + +from topobench.nn.backbones.graph import GREAD +from topobench.nn.wrappers.graph import GNNWrapper + + +class TestGREAD: + """Unit tests for the GREAD backbone.""" + + def setup_method(self): + """Set up test fixtures.""" + torch.manual_seed(0) + self.num_nodes = 8 + self.in_channels = 12 + self.hidden_channels = 16 + self.x = torch.randn(self.num_nodes, self.in_channels) + self.edge_index = torch.randint( + 0, self.num_nodes, (2, self.num_nodes * 2) + ) + + @pytest.mark.parametrize( + "reaction_term", + [ + "bspm", + "fisher", + "allen-cahn", + "zeldovich", + "st", + "fb", + "fb3", + "none", + ], + ) + def test_forward_reaction_terms(self, reaction_term): + """Test the forward pass for every reaction term. + + Parameters + ---------- + reaction_term : str + Reaction term to test. + """ + model = GREAD( + self.in_channels, + self.hidden_channels, + reaction_term=reaction_term, + ) + out = model(self.x, self.edge_index) + assert out.shape == (self.num_nodes, self.hidden_channels) + assert torch.isfinite(out).all() + + def test_invalid_reaction_term(self): + """Test that an invalid reaction term raises a ValueError.""" + with pytest.raises(ValueError): + GREAD(self.in_channels, self.hidden_channels, reaction_term="foo") + + def test_invalid_data_norm(self): + """Test that an invalid data normalization raises a ValueError.""" + with pytest.raises(ValueError): + GREAD(self.in_channels, self.hidden_channels, data_norm="foo") + + def test_gcn_norm(self): + """Test the forward pass with symmetric (GCN) normalization.""" + model = GREAD(self.in_channels, self.hidden_channels, data_norm="gcn") + out = model(self.x, self.edge_index) + assert out.shape == (self.num_nodes, self.hidden_channels) + + def test_beta_diag_and_source(self): + """Test the diagonal beta parameterization and the source term.""" + model = GREAD( + self.in_channels, + self.hidden_channels, + beta_diag=True, + add_source=True, + ) + assert model.b_w.shape == (self.hidden_channels,) + out = model(self.x, self.edge_index) + out.sum().backward() + assert model.b_w.grad is not None + assert model.source_train.grad is not None + + def test_gradient_flow(self): + """Test that gradients flow through alpha, beta and the encoder.""" + model = GREAD(self.in_channels, self.hidden_channels) + out = model(self.x, self.edge_index) + out.sum().backward() + assert model.alpha_train.grad is not None + assert model.beta_train.grad is not None + assert model.encoder.weight.grad is not None + + def test_euler_step_pure_diffusion(self): + """Test one Euler step of pure diffusion against a manual computation. + + With ``reaction_term="none"``, one Euler step computes + ``H(t + tau) = H(t) + tau * sigmoid(alpha) * (A - I) H(t)``. + """ + model = GREAD( + self.in_channels, + self.hidden_channels, + reaction_term="none", + time=1.0, + step_size=1.0, + xn_activation=True, + ) + model.eval() + edge_index, edge_weight = model.normalize_adjacency( + self.edge_index, None, self.num_nodes + ) + with torch.no_grad(): + h0 = model.encoder(self.x) + ah0 = model.spmm(edge_index, edge_weight, self.num_nodes, h0) + alpha = torch.sigmoid(model.alpha_train) + expected = torch.relu(h0 + 1.0 * alpha * (ah0 - h0)) + out = model(self.x, self.edge_index) + assert torch.allclose(out, expected, atol=1e-6) + + def test_rw_normalization_column_stochastic(self): + """Test that the rw-normalized adjacency is column-normalized. + + Matches ``get_rw_adj(..., norm_dim=1)`` of the reference + implementation, where the degree is computed over the target + (column) index. + """ + model = GREAD(self.in_channels, self.hidden_channels) + edge_index, edge_weight = model.normalize_adjacency( + self.edge_index, None, self.num_nodes + ) + col_sums = torch.zeros(self.num_nodes) + col_sums.scatter_add_(0, edge_index[1], edge_weight) + assert torch.allclose(col_sums, torch.ones(self.num_nodes), atol=1e-6) + + def test_beta_diag_weights_reaction_per_channel(self): + """Test the ``beta_diag`` update against a manual Euler step. + + With ``beta_diag=True`` the reaction is scaled by the trainable + diagonal :math:`\\beta_W` — used directly, not through a sigmoid — + instead of the scalar gate :math:`\\beta = \\sigma(\\beta_{train})`, + matching the "(VC)" variants of the paper. + """ + model = GREAD( + self.in_channels, + self.hidden_channels, + reaction_term="fisher", + time=1.0, + step_size=1.0, + beta_diag=True, + xn_activation=False, + ) + model.eval() + edge_index, edge_weight = model.normalize_adjacency( + self.edge_index, None, self.num_nodes + ) + with torch.no_grad(): + h0 = model.encoder(self.x) + ax = model.spmm(edge_index, edge_weight, self.num_nodes, h0) + diffusion = ax - h0 + reaction = -(h0 - 1.0) * h0 + alpha = torch.sigmoid(model.alpha_train) + expected = h0 + (alpha * diffusion + reaction * model.b_w) + out = model(self.x, self.edge_index) + assert torch.allclose(out, expected, atol=1e-6) + + # The per-channel gate must not collapse to the scalar beta path. + scalar_beta = torch.sigmoid(model.beta_train) + scalar_expected = h0 + (alpha * diffusion + scalar_beta * reaction) + assert not torch.allclose(out, scalar_expected, atol=1e-4) + + def test_self_loop_weight(self): + """Test that ``self_loop_weight`` enters the normalized adjacency. + + On the single-edge graph ``0 -> 1`` with self-loop fill value + :math:`w`, the column-stochastic normalization :math:`A D^{-1}` + (degrees over the target index) gives :math:`A_{0,0} = 1`, + :math:`A_{0,1} = 1 / (1 + w)` and :math:`A_{1,1} = w / (1 + w)`. + """ + edge_index = torch.tensor([[0], [1]]) + for weight in (0.5, 2.0): + model = GREAD( + self.in_channels, + self.hidden_channels, + self_loop_weight=weight, + ) + out_index, out_weight = model.normalize_adjacency( + edge_index, None, 2 + ) + dense = torch.zeros(2, 2) + dense[out_index[0], out_index[1]] = out_weight + expected = torch.tensor( + [ + [1.0, 1.0 / (1.0 + weight)], + [0.0, weight / (1.0 + weight)], + ] + ) + assert torch.allclose(dense, expected, atol=1e-6) + + # A non-positive weight must leave the edge set untouched. + model = GREAD( + self.in_channels, self.hidden_channels, self_loop_weight=0.0 + ) + out_index, out_weight = model.normalize_adjacency(edge_index, None, 2) + assert out_index.shape == (2, 1) + assert torch.allclose(out_weight, torch.tensor([1.0]), atol=1e-6) + + def test_dropout_is_training_only(self): + """Test that dropout perturbs training passes but not eval passes.""" + model = GREAD( + self.in_channels, + self.hidden_channels, + input_dropout=0.9, + dropout=0.9, + ) + model.train() + torch.manual_seed(1) + first = model(self.x, self.edge_index) + torch.manual_seed(2) + second = model(self.x, self.edge_index) + assert not torch.allclose(first, second) + + model.eval() + with torch.no_grad(): + repeatable = model(self.x, self.edge_index) + assert torch.allclose( + repeatable, model(self.x, self.edge_index), atol=1e-6 + ) + # In eval mode the result must not depend on the dropout rates. + model.input_dropout = 0.0 + model.dropout = 0.0 + assert torch.allclose( + repeatable, model(self.x, self.edge_index), atol=1e-6 + ) + + def test_euler_step_schedule(self): + """Test the Euler step schedule for divisible and partial horizons.""" + model = GREAD( + self.in_channels, self.hidden_channels, time=3.0, step_size=1.0 + ) + assert model.step_sizes == [1.0, 1.0, 1.0] + model = GREAD( + self.in_channels, self.hidden_channels, time=0.5, step_size=1.0 + ) + assert model.step_sizes == [0.5] + model = GREAD( + self.in_channels, self.hidden_channels, time=2.5, step_size=1.0 + ) + assert model.step_sizes == [1.0, 1.0, 0.5] + + def test_tiny_horizon_step_schedule(self): + """Test that a tiny positive horizon still yields one step.""" + model = GREAD( + self.in_channels, self.hidden_channels, time=1e-12, step_size=1.0 + ) + assert model.step_sizes == [1e-12] + assert sum(model.step_sizes) == 1e-12 + + def test_near_multiple_horizon_step_schedule(self): + """Test that the step sizes sum exactly to a near-multiple horizon.""" + time = 1.0000000005 + model = GREAD( + self.in_channels, self.hidden_channels, time=time, step_size=1.0 + ) + assert len(model.step_sizes) == 2 + assert sum(model.step_sizes) == time + + def test_default_xn_activation_disabled(self): + """Test that the terminal ReLU is disabled by default. + + Matches the ``XN_activation`` CLI default of the reference + implementation. + """ + model = GREAD(self.in_channels, self.hidden_channels) + assert model.xn_activation is False + model.eval() + with torch.no_grad(): + out = model(self.x, self.edge_index) + assert (out < 0).any() + + @pytest.mark.parametrize( + "time, step_size", + [(0.0, 1.0), (-1.0, 1.0), (1.0, 0.0), (1.0, -0.5)], + ) + def test_nonpositive_time_or_step_size_rejected(self, time, step_size): + """Test that non-positive time or step size raises a ValueError. + + Parameters + ---------- + time : float + Terminal integration time. + step_size : float + Euler step size. + """ + with pytest.raises(ValueError): + GREAD( + self.in_channels, + self.hidden_channels, + time=time, + step_size=step_size, + ) + + def test_partial_final_step_numerical(self): + """Test integration with a partial final step against manual Euler. + + With ``reaction_term="none"`` and ``time=2.5``, ``step_size=1.0``, + the model must take steps of 1.0, 1.0 and 0.5: + ``H_{k+1} = H_k + tau_k * sigmoid(alpha) * (A - I) H_k``. + """ + model = GREAD( + self.in_channels, + self.hidden_channels, + reaction_term="none", + time=2.5, + step_size=1.0, + xn_activation=True, + ) + model.eval() + edge_index, edge_weight = model.normalize_adjacency( + self.edge_index, None, self.num_nodes + ) + with torch.no_grad(): + h = model.encoder(self.x) + alpha = torch.sigmoid(model.alpha_train) + for tau in (1.0, 1.0, 0.5): + ah = model.spmm(edge_index, edge_weight, self.num_nodes, h) + h = h + tau * alpha * (ah - h) + expected = torch.relu(h) + out = model(self.x, self.edge_index) + assert torch.allclose(out, expected, atol=1e-6) + + def test_short_horizon_numerical(self): + """Test that ``time < step_size`` takes a single step of size time. + + With ``time=0.5``, ``step_size=1.0`` a single Euler step of size + 0.5 must be taken: + ``H(0.5) = H(0) + 0.5 * sigmoid(alpha) * (A - I) H(0)``. + """ + model = GREAD( + self.in_channels, + self.hidden_channels, + reaction_term="none", + time=0.5, + step_size=1.0, + xn_activation=True, + ) + model.eval() + edge_index, edge_weight = model.normalize_adjacency( + self.edge_index, None, self.num_nodes + ) + with torch.no_grad(): + h0 = model.encoder(self.x) + ah0 = model.spmm(edge_index, edge_weight, self.num_nodes, h0) + alpha = torch.sigmoid(model.alpha_train) + expected = torch.relu(h0 + 0.5 * alpha * (ah0 - h0)) + out = model(self.x, self.edge_index) + assert torch.allclose(out, expected, atol=1e-6) + + def test_zero_initialized_source_contributes_nothing(self): + """Test that a zero-initialized source term contributes zero. + + With ``add_source=True`` and the source coefficient at its zero + initialization, the output must equal that of an identical model + without the source term. + """ + torch.manual_seed(1) + with_source = GREAD( + self.in_channels, + self.hidden_channels, + reaction_term="none", + add_source=True, + ) + torch.manual_seed(1) + without_source = GREAD( + self.in_channels, + self.hidden_channels, + reaction_term="none", + add_source=False, + ) + with_source.eval() + without_source.eval() + assert with_source.source_train.item() == 0.0 + with torch.no_grad(): + out_with = with_source(self.x, self.edge_index) + out_without = without_source(self.x, self.edge_index) + assert torch.allclose(out_with, out_without, atol=1e-7) + + def test_source_term_no_sigmoid(self): + """Test the source-term contribution numerically without sigmoid. + + With a source coefficient ``s``, each Euler step must add + ``tau * s * H(0)`` (not ``tau * sigmoid(s) * H(0)``). + """ + model = GREAD( + self.in_channels, + self.hidden_channels, + reaction_term="none", + time=1.0, + step_size=1.0, + add_source=True, + xn_activation=False, + ) + model.eval() + with torch.no_grad(): + model.source_train.fill_(0.3) + edge_index, edge_weight = model.normalize_adjacency( + self.edge_index, None, self.num_nodes + ) + with torch.no_grad(): + h0 = model.encoder(self.x) + ah0 = model.spmm(edge_index, edge_weight, self.num_nodes, h0) + alpha = torch.sigmoid(model.alpha_train) + expected = h0 + 1.0 * (alpha * (ah0 - h0) + 0.3 * h0) + out = model(self.x, self.edge_index) + assert torch.allclose(out, expected, atol=1e-6) + + def test_xn_activation_toggle(self): + """Test the terminal ReLU toggle numerically. + + With ``xn_activation=False`` the raw terminal state is returned; + with ``xn_activation=True`` its ReLU is returned. + """ + torch.manual_seed(2) + relu_model = GREAD( + self.in_channels, + self.hidden_channels, + xn_activation=True, + ) + torch.manual_seed(2) + raw_model = GREAD( + self.in_channels, + self.hidden_channels, + xn_activation=False, + ) + relu_model.eval() + raw_model.eval() + with torch.no_grad(): + out_relu = relu_model(self.x, self.edge_index) + out_raw = raw_model(self.x, self.edge_index) + assert (out_raw < 0).any() + assert torch.allclose(out_relu, torch.relu(out_raw), atol=1e-7) + + @pytest.mark.parametrize( + "reaction_term, reaction_fn", + [ + ("fisher", lambda h, h0: -(h - 1) * h), + ("allen-cahn", lambda h, h0: -(h**2 - 1) * h), + ("zeldovich", lambda h, h0: -(h**2 - h) * h), + ("st", lambda h, h0: h0), + ], + ) + def test_pointwise_reactions_numerical(self, reaction_term, reaction_fn): + """Test pointwise reaction terms against manual Euler computations. + + One Euler step computes ``H1 = H0 + tau * (sigmoid(alpha) * + (A - I) H0 + sigmoid(beta) * r(H0))``. + + Parameters + ---------- + reaction_term : str + Reaction term to test. + reaction_fn : callable + Manual implementation of the reaction term. + """ + model = GREAD( + self.in_channels, + self.hidden_channels, + reaction_term=reaction_term, + time=1.0, + step_size=1.0, + xn_activation=False, + ) + model.eval() + edge_index, edge_weight = model.normalize_adjacency( + self.edge_index, None, self.num_nodes + ) + with torch.no_grad(): + h0 = model.encoder(self.x) + ah0 = model.spmm(edge_index, edge_weight, self.num_nodes, h0) + alpha = torch.sigmoid(model.alpha_train) + beta = torch.sigmoid(model.beta_train) + expected = h0 + 1.0 * ( + alpha * (ah0 - h0) + beta * reaction_fn(h0, h0) + ) + out = model(self.x, self.edge_index) + assert torch.allclose(out, expected, atol=1e-6) + + @pytest.mark.parametrize("reaction_term", ["bspm", "fb", "fb3"]) + def test_graph_reactions_numerical(self, reaction_term): + """Test graph-coupled reaction terms against manual computations. + + Covers ``bspm`` (``-A(A - I)H``), ``fb`` (``H + AH``) and + ``fb3`` (``fb`` plus an extra ``H`` inside the beta gate). + + Parameters + ---------- + reaction_term : str + Reaction term to test. + """ + model = GREAD( + self.in_channels, + self.hidden_channels, + reaction_term=reaction_term, + time=1.0, + step_size=1.0, + xn_activation=False, + ) + model.eval() + edge_index, edge_weight = model.normalize_adjacency( + self.edge_index, None, self.num_nodes + ) + with torch.no_grad(): + h0 = model.encoder(self.x) + ah0 = model.spmm(edge_index, edge_weight, self.num_nodes, h0) + diffusion = ah0 - h0 + alpha = torch.sigmoid(model.alpha_train) + beta = torch.sigmoid(model.beta_train) + if reaction_term == "bspm": + reaction = -model.spmm( + edge_index, edge_weight, self.num_nodes, diffusion + ) + f = alpha * diffusion + beta * reaction + else: + reaction = h0 + model.spmm( + edge_index, edge_weight, self.num_nodes, h0 + ) + if reaction_term == "fb3": + f = alpha * diffusion + beta * (reaction + h0) + else: + f = alpha * diffusion + beta * reaction + expected = h0 + 1.0 * f + out = model(self.x, self.edge_index) + assert torch.allclose(out, expected, atol=1e-6) + + def test_edge_weight_passthrough(self): + """Test the forward pass with explicit edge weights.""" + model = GREAD(self.in_channels, self.hidden_channels) + edge_weight = torch.rand(self.edge_index.size(1)) + out = model(self.x, self.edge_index, edge_weight=edge_weight) + assert out.shape == (self.num_nodes, self.hidden_channels) + + def test_with_gnn_wrapper(self, random_graph_input): + """Test GREAD within the GNNWrapper. + + Parameters + ---------- + random_graph_input : tuple + Fixture with random graph inputs. + """ + x, _, _, edges_1, _ = random_graph_input + batch = torch_geometric.data.Data( + x_0=x, + y=x, + x=x, + edge_index=edges_1, + batch_0=torch.zeros(x.shape[0], dtype=torch.long), + ) + model = GREAD(x.shape[1], x.shape[1]) + wrapper = GNNWrapper( + model, + out_channels=x.shape[1], + num_cell_dimensions=1, + ) + model_out = wrapper(batch) + assert model_out["x_0"].shape == x.shape diff --git a/test/pipeline/test_pipeline.py b/test/pipeline/test_pipeline.py index a61165ae9..1b0966cd3 100644 --- a/test/pipeline/test_pipeline.py +++ b/test/pipeline/test_pipeline.py @@ -7,7 +7,13 @@ DATASET = "graph/MUTAG" # ADD YOUR DATASET HERE -MODELS = ["graph/gcn", "cell/topotune", "simplicial/topotune"] # ADD ONE OR SEVERAL MODELS +MODELS = [ + "graph/gread", + "graph/gread_fisher", + "graph/gread_allen_cahn", + "graph/gread_zeldovich", + "graph/gread_source_term", +] # ADD ONE OR SEVERAL MODELS class TestPipeline: diff --git a/topobench/nn/backbones/graph/gread.py b/topobench/nn/backbones/graph/gread.py new file mode 100644 index 000000000..8f2c12a26 --- /dev/null +++ b/topobench/nn/backbones/graph/gread.py @@ -0,0 +1,427 @@ +"""GREAD: Graph Neural Reaction-Diffusion Networks. + +Implementation of the paper "GREAD: Graph Neural Reaction-Diffusion +Networks" (Choi et al., ICML 2023, https://arxiv.org/abs/2211.14208), +adapted from the official reference implementation +https://github.com/jeongwhanchoi/GREAD. +""" + +import torch +import torch.nn.functional as F +from torch import nn +from torch_geometric.utils import ( + add_remaining_self_loops, + scatter, +) + + +class GREAD(nn.Module): + r"""GREAD: Graph Neural Reaction-Diffusion Network backbone. + + GREAD (Choi et al., ICML 2023, https://arxiv.org/abs/2211.14208) + evolves node representations :math:`\mathbf{H}(t)` with the + reaction-diffusion equation (Section 3.1, Eqs. (6)-(8) of the paper): + + .. math:: + \mathbf{f}(\mathbf{H}(t)) := \frac{d\mathbf{H}(t)}{dt} = + -\alpha \, \mathbf{L}\mathbf{H}(t) + + \beta \, \mathbf{r}(\mathbf{H}(t)), + + with initial value :math:`\mathbf{H}(0) = \mathbf{e}(\mathbf{X})` + (Eq. (6)), where :math:`\mathbf{e}` is an encoding layer, + :math:`\mathbf{L} = \mathbf{I} - \mathbf{A}` the Laplacian of a + normalized adjacency matrix :math:`\mathbf{A}`, and + :math:`\mathbf{r}` one of the reaction terms of Eq. (10) of the + paper: + + - ``"bspm"``: blurring-sharpening + :math:`(\mathbf{A} - \mathbf{A}^2)\mathbf{H}(t)` + (GREAD-BS, Eq. (14)). + - ``"fisher"``: Fisher + :math:`\mathbf{H}(t)\odot(1 - \mathbf{H}(t))` (GREAD-F). + - ``"allen-cahn"``: Allen-Cahn + :math:`\mathbf{H}(t)\odot(1 - \mathbf{H}(t)^{\circ 2})` + (GREAD-AC). + - ``"zeldovich"``: Zeldovich + :math:`\mathbf{H}(t)\odot(\mathbf{H}(t) - + \mathbf{H}(t)^{\circ 2})` (GREAD-Z). + - ``"st"``: source term :math:`\mathbf{H}(0)` (GREAD-ST). + - ``"fb"``: filter bank (high-pass) reaction (GREAD-FB). Eq. (10) + of the paper writes it as :math:`\mathbf{L}\mathbf{H}(t)`, while + the reference implementation computes + :math:`(\mathbf{I} + \mathbf{A})\mathbf{H}(t)`; this + implementation follows the reference code. + - ``"fb3"``: filter bank with identity channel (GREAD-FB*), the + ``"fb"`` reaction plus :math:`\mathbf{H}(t)`, again following the + reference code. + - ``"none"``: no reaction, i.e. pure diffusion. + + The ODE is integrated with the explicit Euler method, the base + solver in which Eq. (5) of the paper is written and one of the two + fixed-grid solvers (``euler`` / ``rk4``) selected per dataset by + the reference implementation + (https://github.com/jeongwhanchoi/GREAD, + ``src/gread_params.py``). Integration reaches the terminal time + ``time`` up to floating-point rounding: full steps of size + ``step_size`` are taken and, when ``time`` is not an integer + multiple of ``step_size``, a final shorter step covers the remaining + duration. + + The scalar gates are parameterized as + :math:`\alpha = \sigma(\alpha_{\text{train}})` and + :math:`\beta = \sigma(\beta_{\text{train}})` following + ``ODEFuncGread.forward`` of the reference code; with + ``beta_diag=True`` the reaction is instead weighted by a trainable + diagonal matrix :math:`\beta_W` (the "(VC)" variants in the paper). + + Parameters + ---------- + in_channels : int + Number of input features. + hidden_channels : int + Number of hidden units. + reaction_term : str, optional + Reaction term, one of ``"bspm"``, ``"fisher"``, ``"allen-cahn"``, + ``"zeldovich"``, ``"st"``, ``"fb"``, ``"fb3"``, ``"none"`` + (default: ``"bspm"``). + time : float, optional + Terminal integration time :math:`T`; must be positive + (default: 3.0). + step_size : float, optional + Step size :math:`\tau` of the explicit Euler solver; must be + positive (default: 1.0). + beta_diag : bool, optional + If True, use a trainable diagonal matrix ("vector channel") + instead of the scalar :math:`\beta` gate (default: False). + add_source : bool, optional + If True, add a learnable source term + :math:`s\,\mathbf{H}(0)` to the dynamics, with :math:`s` + initialized at zero so the source contributes nothing until + trained (default: False). + data_norm : str, optional + Adjacency normalization: ``"rw"`` for the column-stochastic + random-walk normalization :math:`\mathbf{A}\mathbf{D}^{-1}` + (degrees taken over the target index, matching + ``get_rw_adj(..., norm_dim=1)`` of the reference + implementation) or ``"gcn"`` for the symmetric normalization + :math:`\mathbf{D}^{-1/2}\mathbf{A}\mathbf{D}^{-1/2}` + (default: ``"rw"``). + self_loop_weight : float, optional + Weight of the self loops added to the adjacency matrix before + normalization (default: 1.0). + xn_activation : bool, optional + If True, apply a terminal ReLU to the integrated + representations :math:`\mathbf{H}(T)` before the output + dropout (``XN_activation`` in the reference implementation, + whose CLI default is ``False``; it is enabled only for selected + dataset configurations). Default: ``False``. + input_dropout : float, optional + Dropout rate applied to the input features (default: 0.0). + dropout : float, optional + Dropout rate applied to the final representation + (default: 0.0). + **kwargs + Additional arguments (ignored). + """ + + _REACTION_TERMS = ( + "bspm", + "fisher", + "allen-cahn", + "zeldovich", + "st", + "fb", + "fb3", + "none", + ) + + def __init__( + self, + in_channels, + hidden_channels, + reaction_term="bspm", + time=3.0, + step_size=1.0, + beta_diag=False, + add_source=False, + data_norm="rw", + self_loop_weight=1.0, + xn_activation=False, + input_dropout=0.0, + dropout=0.0, + **kwargs, + ): + super().__init__() + if reaction_term not in self._REACTION_TERMS: + raise ValueError( + f"Unknown reaction term '{reaction_term}'. " + f"Expected one of {self._REACTION_TERMS}." + ) + if data_norm not in ("rw", "gcn"): + raise ValueError( + f"Unknown data normalization '{data_norm}'. " + "Expected 'rw' or 'gcn'." + ) + if time <= 0: + raise ValueError( + f"Terminal time must be positive, got time={time}." + ) + if step_size <= 0: + raise ValueError( + f"Step size must be positive, got step_size={step_size}." + ) + self.in_channels = in_channels + self.hidden_channels = hidden_channels + self.out_channels = hidden_channels + self.reaction_term = reaction_term + self.time = time + self.step_size = step_size + self.step_sizes = self._euler_step_sizes(time, step_size) + self.beta_diag = beta_diag + self.add_source = add_source + self.data_norm = data_norm + self.self_loop_weight = self_loop_weight + self.xn_activation = xn_activation + self.input_dropout = input_dropout + self.dropout = dropout + + # Encoding layer E (m1 in the reference implementation). + self.encoder = nn.Linear(in_channels, hidden_channels) + + # Scalar gates alpha and beta (sigmoid-activated, initialized at + # zero as in ODEFunc.__init__ of the reference implementation). + self.alpha_train = nn.Parameter(torch.tensor(0.0)) + self.beta_train = nn.Parameter(torch.tensor(0.0)) + if beta_diag: + self.b_w = nn.Parameter(torch.empty(hidden_channels)) + if add_source: + self.source_train = nn.Parameter(torch.tensor(0.0)) + + self.reset_parameters() + + @staticmethod + def _euler_step_sizes(time, step_size): + """Compute the explicit Euler step schedule. + + Full steps of ``step_size`` are taken while they fit within the + horizon; a final shorter step covers any remaining duration so + that the step sizes sum to ``time`` up to floating-point + rounding. At least one step is always taken for a positive + horizon. + + Parameters + ---------- + time : float + Terminal integration time. + step_size : float + Nominal step size. + + Returns + ------- + list of float + The sizes of the successive Euler steps; they sum to + ``time`` up to floating-point rounding. + """ + n_full = int(time / step_size) + remainder = time - n_full * step_size + sizes = [step_size] * n_full + if remainder > 0: + sizes.append(remainder) + return sizes + + def reset_parameters(self): + """Reset the learnable parameters.""" + self.encoder.reset_parameters() + with torch.no_grad(): + self.alpha_train.zero_() + self.beta_train.zero_() + if self.beta_diag: + nn.init.uniform_(self.b_w, a=-1.0, b=1.0) + if self.add_source: + with torch.no_grad(): + self.source_train.zero_() + + def normalize_adjacency(self, edge_index, edge_weight, num_nodes): + r"""Normalize the adjacency matrix. + + Adds weighted self loops and normalizes the adjacency matrix, + following ``get_rw_adj`` (``data_norm="rw"``) and + ``gcn_norm_fill_val`` (``data_norm="gcn"``) of the reference + implementation (``src/utils.py``). + + Parameters + ---------- + edge_index : torch.Tensor + Edge index of shape ``[2, num_edges]``. + edge_weight : torch.Tensor or None + Optional edge weights of shape ``[num_edges]``. + num_nodes : int + Number of nodes. + + Returns + ------- + tuple of (torch.Tensor, torch.Tensor) + Normalized edge index and edge weights. + """ + if edge_weight is None: + edge_weight = torch.ones( + edge_index.size(1), + dtype=torch.float, + device=edge_index.device, + ) + if self.self_loop_weight > 0: + edge_index, edge_weight = add_remaining_self_loops( + edge_index, + edge_weight, + fill_value=self.self_loop_weight, + num_nodes=num_nodes, + ) + row, col = edge_index[0], edge_index[1] + if self.data_norm == "rw": + # A D^{-1}: degrees over the target (column) index, i.e. a + # column-stochastic operator (norm_dim=1 in the reference + # implementation). + deg = scatter( + edge_weight, col, dim=0, dim_size=num_nodes, reduce="sum" + ) + deg_inv = deg.pow(-1) + deg_inv.masked_fill_(deg_inv == float("inf"), 0) + edge_weight = deg_inv[col] * edge_weight + else: + deg = scatter( + edge_weight, col, dim=0, dim_size=num_nodes, reduce="sum" + ) + deg_inv_sqrt = deg.pow(-0.5) + deg_inv_sqrt.masked_fill_(deg_inv_sqrt == float("inf"), 0) + edge_weight = deg_inv_sqrt[row] * edge_weight * deg_inv_sqrt[col] + return edge_index, edge_weight + + @staticmethod + def spmm(edge_index, edge_weight, num_nodes, x): + r"""Multiply the sparse adjacency matrix with a dense matrix. + + Computes :math:`\mathbf{A}\mathbf{X}`, the equivalent of + ``ODEFuncGread.sparse_multiply`` of the reference implementation. + + Parameters + ---------- + edge_index : torch.Tensor + Edge index of shape ``[2, num_edges]``. + edge_weight : torch.Tensor + Edge weights of shape ``[num_edges]``. + num_nodes : int + Number of nodes. + x : torch.Tensor + Dense feature matrix of shape ``[num_nodes, channels]``. + + Returns + ------- + torch.Tensor + The product :math:`\mathbf{A}\mathbf{X}`. + """ + row, col = edge_index[0], edge_index[1] + return scatter( + edge_weight.unsqueeze(-1) * x[col], + row, + dim=0, + dim_size=num_nodes, + reduce="sum", + ) + + def reaction(self, x, diffusion, x0, edge_index, edge_weight, num_nodes): + r"""Compute the reaction term :math:`\mathbf{r}(\mathbf{H}, \mathbf{A})`. + + Implements the reaction terms of Eq. (10) of the paper, matching + ``ODEFuncGread.forward`` of the reference implementation. + + Parameters + ---------- + x : torch.Tensor + Current node representations :math:`\mathbf{H}(t)`. + diffusion : torch.Tensor + Diffusion term :math:`(\mathbf{A} - \mathbf{I})\mathbf{H}(t)`. + x0 : torch.Tensor + Initial node representations :math:`\mathbf{H}(0)`. + edge_index : torch.Tensor + Edge index of shape ``[2, num_edges]``. + edge_weight : torch.Tensor + Normalized edge weights of shape ``[num_edges]``. + num_nodes : int + Number of nodes. + + Returns + ------- + torch.Tensor + The reaction term. + """ + if self.reaction_term == "bspm": + return -self.spmm(edge_index, edge_weight, num_nodes, diffusion) + if self.reaction_term == "fisher": + return -(x - 1) * x + if self.reaction_term == "allen-cahn": + return -(x**2 - 1) * x + if self.reaction_term == "zeldovich": + return -(x**2 - x) * x + if self.reaction_term == "st": + return x0 + if self.reaction_term in ("fb", "fb3"): + ax = -self.spmm(edge_index, edge_weight, num_nodes, x) + return x - ax + return torch.zeros_like(x) + + def forward(self, x, edge_index, batch=None, edge_weight=None): + r"""Forward pass. + + Encodes the input features, integrates the reaction-diffusion + equation with the explicit Euler method and returns the terminal + node representations :math:`\mathbf{H}(T)`. + + Parameters + ---------- + x : torch.Tensor + Input node features of shape ``[num_nodes, in_channels]``. + edge_index : torch.Tensor + Edge index of shape ``[2, num_edges]``. + batch : torch.Tensor, optional + Batch assignment vector (unused, present for API + compatibility). + edge_weight : torch.Tensor, optional + Optional edge weights of shape ``[num_edges]``. + + Returns + ------- + torch.Tensor + Node representations of shape + ``[num_nodes, hidden_channels]``. + """ + num_nodes = x.size(0) + edge_index, edge_weight = self.normalize_adjacency( + edge_index, edge_weight, num_nodes + ) + + x = F.dropout(x, self.input_dropout, training=self.training) + x = self.encoder(x) + x0 = x + + alpha = torch.sigmoid(self.alpha_train) + beta = torch.sigmoid(self.beta_train) + + for tau in self.step_sizes: + ax = self.spmm(edge_index, edge_weight, num_nodes, x) + diffusion = ax - x + reaction = self.reaction( + x, diffusion, x0, edge_index, edge_weight, num_nodes + ) + if self.beta_diag: + f = alpha * diffusion + reaction * self.b_w + elif self.reaction_term == "fb3": + f = alpha * diffusion + beta * (reaction + x) + else: + f = alpha * diffusion + beta * reaction + if self.add_source: + f = f + self.source_train * x0 + x = x + tau * f + + if self.xn_activation: + x = F.relu(x) + return F.dropout(x, self.dropout, training=self.training)