Add UI network inspectors

README.md

@@ -52,6 +52,23 @@ Necessary dependency should be installed automatically in the process.
 
 Check the basic tutorial [here](docs/basic_tutorial/basic_tutorial_lan_torch.ipynb).
 
+### Network Inspectors UI
+
+LANfactory includes a Streamlit interface for interactive network inspection
+workflows (KDE vs LAN likelihoods and LAN manifold plots).
+
+Install the UI dependencies:
+
+```bash
+uv sync --extra ui
+```
+
+Launch the app:
+
+```bash
+uv run network-inspectors-ui
+```
+
 ### Command Line Interface
 
 LANfactory includes a command line interface with the commands `jaxtrain` and `torchtrain`, which train neural networks using `jax` and `torch` as backends, respectively.
@@ -133,6 +150,30 @@ To make your own configuration file, you can copy the example above into a new `
 
 If you are using `uv`, you can also use the `uv run` command to run `jaxtrain` or `torchtrain` from the command line
 
+### Batch Training Multiple Torch Models
+
+To generate multiple torch models (for example, `angle` and `ddm`) in one command,
+use the helper script:
+
+```sh
+scripts/train_torch_models_batch.sh \
+  --training-data-base data/data \
+  --networks-path-base data/torch_models \
+  --models angle,ddm \
+  --network-ids 0,1,2
+```
+
+If all models should use the same training-data folder, use:
+
+```sh
+scripts/train_torch_models_batch.sh \
+  --training-data-folder data/data/angle \
+  --models angle \
+  --network-ids 0,1
+```
+
+You can validate setup without training via `--dry-run`.
+
 ### TorchMLP to ONNX Converter
 
 Once you have trained your model, you can convert it to the ONNX format using the provided `transform-onnx` command.

pyproject.toml

@@ -57,13 +57,15 @@ mlflow = ["mlflow>=3.14.0"]
 hf = ["huggingface-hub>=0.20.0"]
 sbi = ["sbi>=0.26", "nflows>=0.14"]
 bayesflow = ["bayesflow>=2.0.8", "keras>=3.12"]
+ui = ["streamlit>=1.40.0"]
 all = [
     "mlflow>=3.14.0",
     "huggingface-hub>=0.20.0",
     "sbi>=0.26",
     "nflows>=0.14",
     "bayesflow>=2.0.8",
     "keras>=3.12",
+    "streamlit>=1.40.0",
 ]
 
 [dependency-groups]
@@ -164,6 +166,8 @@ torchtrain = "lanfactory.cli.torch_train:app"
 transform-onnx = "lanfactory.onnx.transform_onnx:app"
 upload-hf = "lanfactory.cli.upload_hf:app"
 download-hf = "lanfactory.cli.download_hf:app"
+network-inspectors-ui = "lanfactory.cli.network_inspectors_ui:app"
 
 [tool.setuptools.package-data]
 "lanfactory.cli" = ["config_network_training_lan.yaml"]
+"lanfactory.network_inspectors" = ["styles.css"]

scripts/train_torch_models_batch.sh

@@ -0,0 +1,294 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+# Train multiple torch LAN models (for example angle and ddm) in one command.
+# Uses uv + torchtrain and writes outputs into the chosen networks path.
+
+CONFIG_TEMPLATE="src/lanfactory/cli/config_network_training_lan.yaml"
+TRAINING_DATA_FOLDER=""
+TRAINING_DATA_BASE=""
+NETWORKS_PATH_BASE="data/torch_models"
+MODELS="angle,ddm"
+NETWORK_IDS="0,1,2"
+DL_WORKERS="1"
+LOG_LEVEL="INFO"
+DRY_RUN="0"
+
+usage() {
+    cat <<'EOF'
+Usage:
+  scripts/train_torch_models_batch.sh [options]
+
+Options:
+  --config-template PATH       YAML template with a MODEL field.
+                               Default: src/lanfactory/cli/config_network_training_lan.yaml
+  --training-data-folder PATH  Shared training data folder used for all models.
+  --training-data-base PATH    Base folder where each model has its own subfolder.
+                               Example: <base>/angle and <base>/ddm
+  --networks-path-base PATH    Base output path for trained models.
+                               Default: data/torch_models
+  --models CSV                 Comma-separated model names.
+                               Default: angle,ddm
+  --network-ids CSV            Comma-separated network IDs.
+                               Default: 0,1,2
+  --dl-workers N               DataLoader workers passed to torchtrain.
+                               Default: 1
+  --log-level LEVEL            Logging level for torchtrain.
+                               Default: INFO
+  --dry-run                    Validate commands without training.
+  --help                       Show this help message.
+
+Examples:
+  scripts/train_torch_models_batch.sh \
+    --training-data-base data/data \
+    --networks-path-base data/torch_models \
+    --models angle,ddm \
+    --network-ids 0,1,2,3
+
+  scripts/train_torch_models_batch.sh \
+    --training-data-folder data/data/angle \
+    --models angle \
+    --network-ids 0,1
+EOF
+}
+
+while [[ $# -gt 0 ]]; do
+    case "$1" in
+        --config-template)
+            CONFIG_TEMPLATE="$2"
+            shift 2
+            ;;
+        --training-data-folder)
+            TRAINING_DATA_FOLDER="$2"
+            shift 2
+            ;;
+        --training-data-base)
+            TRAINING_DATA_BASE="$2"
+            shift 2
+            ;;
+        --networks-path-base)
+            NETWORKS_PATH_BASE="$2"
+            shift 2
+            ;;
+        --models)
+            MODELS="$2"
+            shift 2
+            ;;
+        --network-ids)
+            NETWORK_IDS="$2"
+            shift 2
+            ;;
+        --dl-workers)
+            DL_WORKERS="$2"
+            shift 2
+            ;;
+        --log-level)
+            LOG_LEVEL="$2"
+            shift 2
+            ;;
+        --dry-run)
+            DRY_RUN="1"
+            shift
+            ;;
+        --help)
+            usage
+            exit 0
+            ;;
+        *)
+            echo "Unknown option: $1" >&2
+            usage
+            exit 1
+            ;;
+    esac
+done
+
+if [[ ! -f "$CONFIG_TEMPLATE" ]]; then
+    echo "Config template not found: $CONFIG_TEMPLATE" >&2
+    exit 1
+fi
+
+if [[ -z "$TRAINING_DATA_FOLDER" && -z "$TRAINING_DATA_BASE" ]]; then
+    echo "Provide either --training-data-folder or --training-data-base" >&2
+    exit 1
+fi
+
+IFS=',' read -r -a MODELS_ARR <<< "$MODELS"
+IFS=',' read -r -a IDS_ARR <<< "$NETWORK_IDS"
+
+TMP_DIR="$(mktemp -d)"
+trap 'rm -rf "$TMP_DIR"' EXIT
+
+make_model_config() {
+    local template="$1"
+    local model="$2"
+    local out_file="$3"
+
+    # Replace MODEL: ... in YAML while preserving all other lines.
+    awk -v model="$model" '
+        BEGIN { replaced = 0 }
+        /^[[:space:]]*MODEL[[:space:]]*:/ && replaced == 0 {
+            print "MODEL: \"" model "\""
+            replaced = 1
+            next
+        }
+        { print }
+        END {
+            if (replaced == 0) {
+                print "MODEL: \"" model "\""
+            }
+        }
+    ' "$template" > "$out_file"
+}
+
+resolve_model_training_folder() {
+    local candidate_folder="$1"
+    local model="$2"
+
+    if [[ ! -d "$candidate_folder" ]]; then
+        echo "$candidate_folder"
+        return
+    fi
+
+    # Preferred: pickle shards directly in the provided folder.
+    if find "$candidate_folder" -maxdepth 1 -type f -name '*.pickle' | grep -q .; then
+        echo "$candidate_folder"
+        return
+    fi
+
+    # Common layout: one subfolder per model.
+    if [[ -d "$candidate_folder/$model" ]] && find "$candidate_folder/$model" -maxdepth 1 -type f -name '*.pickle' | grep -q .; then
+        echo "$candidate_folder/$model"
+        return
+    fi
+
+    # Fallback: if exactly one immediate subfolder contains pickle shards, use it.
+    local subdir_count
+    subdir_count="$(find "$candidate_folder" -mindepth 1 -maxdepth 1 -type d | wc -l | tr -d ' ')"
+    if [[ "$subdir_count" == "1" ]]; then
+        local only_subdir
+        only_subdir="$(find "$candidate_folder" -mindepth 1 -maxdepth 1 -type d | head -n 1)"
+        if find "$only_subdir" -maxdepth 1 -type f -name '*.pickle' | grep -q .; then
+            echo "$only_subdir"
+            return
+        fi
+    fi
+
+    echo "$candidate_folder"
+}
+
+validate_training_folder() {
+    local folder="$1"
+    local model="$2"
+
+    local n_pickles
+    n_pickles="$(find "$folder" -maxdepth 1 -type f -name '*.pickle' | wc -l | tr -d ' ')"
+
+    if [[ "$n_pickles" == "0" ]]; then
+        echo "No .pickle files found for model '$model' in: $folder" >&2
+        echo "Expected either:" >&2
+        echo "  1) <folder>/*.pickle" >&2
+        echo "  2) <folder>/$model/*.pickle" >&2
+        exit 1
+    fi
+
+    local validation_out
+    if ! validation_out="$((uv run python - "$folder" <<'PY'
+import glob
+import os
+import pickle
+import sys
+
+folder = sys.argv[1]
+files = sorted(glob.glob(os.path.join(folder, "*.pickle")))
+if not files:
+    print("ERR_NO_PICKLES")
+    raise SystemExit(2)
+
+with open(files[0], "rb") as f:
+    obj = pickle.load(f)
+
+if not isinstance(obj, dict):
+    print("ERR_NOT_DICT")
+    raise SystemExit(3)
+
+keys = set(obj.keys())
+required = {"lan_data", "lan_labels"}
+if not required.issubset(keys):
+    print("ERR_BAD_KEYS")
+    print(",".join(sorted(keys)))
+    raise SystemExit(4)
+
+print("OK_KEYS")
+PY
+))"; then
+        if grep -q "ERR_BAD_KEYS" <<<"$validation_out"; then
+            echo "Training data format mismatch for model '$model' in: $folder" >&2
+            echo "Expected pickle keys: lan_data and lan_labels" >&2
+            echo "Found keys:" >&2
+            echo "$(tail -n 1 <<<"$validation_out")" >&2
+            exit 1
+        fi
+
+        if grep -q "ERR_NOT_DICT" <<<"$validation_out"; then
+            echo "Unexpected pickle format in: $folder" >&2
+            echo "Expected each shard to be a dict with lan_data/lan_labels." >&2
+            exit 1
+        fi
+
+        echo "Failed to validate training folder '$folder'." >&2
+        echo "$validation_out" >&2
+        exit 1
+    fi
+}
+
+for model in "${MODELS_ARR[@]}"; do
+    model="${model//[[:space:]]/}"
+    if [[ -z "$model" ]]; then
+        continue
+    fi
+
+    if [[ -n "$TRAINING_DATA_FOLDER" ]]; then
+        model_training_folder="$(resolve_model_training_folder "$TRAINING_DATA_FOLDER" "$model")"
+    else
+        model_training_folder="$(resolve_model_training_folder "$TRAINING_DATA_BASE/$model" "$model")"
+    fi
+
+    if [[ ! -d "$model_training_folder" ]]; then
+        echo "Training data folder not found for model '$model': $model_training_folder" >&2
+        exit 1
+    fi
+
+    validate_training_folder "$model_training_folder" "$model"
+
+    model_cfg="$TMP_DIR/config_${model}.yaml"
+    make_model_config "$CONFIG_TEMPLATE" "$model" "$model_cfg"
+
+    for net_id in "${IDS_ARR[@]}"; do
+        net_id="${net_id//[[:space:]]/}"
+        if [[ -z "$net_id" ]]; then
+            continue
+        fi
+
+        cmd=(
+            uv run torchtrain
+            --config-path "$model_cfg"
+            --training-data-folder "$model_training_folder"
+            --networks-path-base "$NETWORKS_PATH_BASE"
+            --network-id "$net_id"
+            --dl-workers "$DL_WORKERS"
+            --log-level "$LOG_LEVEL"
+        )
+
+        if [[ "$DRY_RUN" == "1" ]]; then
+            cmd+=(--dry-run)
+        fi
+
+        echo ""
+        echo "=== Training model=$model network_id=$net_id ==="
+        echo "Training data: $model_training_folder"
+        "${cmd[@]}"
+    done
+done
+
+echo ""
+echo "Batch training complete."

src/lanfactory/cli/network_inspectors_ui.py

@@ -0,0 +1,23 @@
+"""Launcher for the network inspectors Streamlit app."""
+
+from __future__ import annotations
+
+from pathlib import Path
+import sys
+
+
+def app() -> None:
+    """Start the LANfactory network inspectors UI via Streamlit."""
+    try:
+        from streamlit.web import cli as stcli
+    except ImportError as exc:  # pragma: no cover
+        raise ImportError(
+            "Streamlit is required for the network-inspectors-ui command. "
+            "Install optional dependencies with: uv sync --extra ui"
+        ) from exc
+
+    app_path = (
+        Path(__file__).resolve().parents[1] / "network_inspectors" / "streamlit_app.py"
+    )
+    sys.argv = ["streamlit", "run", str(app_path)]
+    raise SystemExit(stcli.main())