fix/tree-node-distribution (#6)

- created viewer.py (e20e8e9ca6d96ec1b1fe0fb7253cdca087068baf)
- modified sidebar (6793fc46233a6dedbb092fc8aa2a2794ae0977e1)
- updated graph visualization with pyvis (a8d32863cd73fde11d669011569da5500f65c08b)
- implemented graph style selection from PyVis and NetworkX (1fdf3d016474aa2c2a769a27c24afec096980e79)
- moved sample details inside expander (5c3f970c0f9f1ecee10a9a4c78ba984ab7921832)
- modified graph height for NetworkX (191ea1806d190eb182645c0d76d20a286b413793)
- added number of samples (83dd85734e43f1f0bc8e9c65e951ca622c4398d2)
- updated README.md (68569e789bc27e269d5263987529404acdcad276)
- refactored main() into two functions (6d690044f3e789e99dfbf6390551c283b3edf2e1)
- added random pick button (8c25bcecd39c01b4229c6ef88c01f69ae9109444)
- renamed a function (c0886ffd3913f1425e469f9a999615cdd74609d1)
- debugged number of nodes for tree base graphs (dab7f7a0b0197430f67c31c2c6e18c4071748d1b)
- re-generated tree_cycle (9d6b82e8ca50013eff6485b0c78635dfc1347518)
- re-generated tree_grid (4f762c4453b94991b0d0d408ced4136c3b48a158)
- updated README (5db1de77ce5bb092aa087338c808fd1f170e5e9a)

README.md

@@ -108,9 +108,9 @@ Node counts depend on the underlying base generator and motif size:
 
 - `ba_shapes` (Barabási–Albert, `num_edges = 2`): 8–20 nodes.
   ![BA shapes node counts](fig/ba_shapes/node_count_hist.svg)
-- `tree_cycle` (Prüfer trees + 6-cycle motif): 8–21 nodes.
+- `tree_cycle` (Prüfer trees + 6-cycle motif): 10–20 nodes.
   ![Tree cycle node counts](fig/tree_cycle/node_count_hist.svg)
-- `tree_grid` (Prüfer trees + 3×3 grid motif): 8–24 nodes.
+- `tree_grid` (Prüfer trees + 3×3 grid motif): 10–20 nodes.
   ![Tree grid node counts](fig/tree_grid/node_count_hist.svg)
 - `ba_two_motifs` (Barabási–Albert; house vs 5-cycle motifs): 8–20 nodes.
   ![BA two motifs node counts](fig/ba_two_motifs/node_count_hist.svg)
@@ -123,7 +123,8 @@ Observed frequencies across the full 2,000-example splits for each subset are sh
 
 All examples are synthetically generated using [PyTorch Geometric](https://pytorch-geometric.readthedocs.io/) (2.7.0) primitives:
 
-- Detection subsets: positive samples start from certain base graphs whose node counts are drawn from a uniform distribution before a motif is injected, yielding final graphs with 8–20 nodes. Negative samples use the same process with node counts drawn from `range(MIN_NODES, MAX_NODES + 1)` (default 8–20) and no motif.
+- Detection subsets: positive samples start from certain base graphs whose node counts are drawn from a uniform distribution before a motif is injected, yielding final graphs with 8–20 or 10-20 nodes depending on the type of base graphs (BA or tree).
+  Negative samples use the same process with node counts drawn from `range(MIN_NODES, MAX_NODES + 1)` and no motif.
 - `ba_two_motifs`: both classes contain motifs injected into Barabási–Albert graphs—house motifs vs 5-cycle motifs—with final graphs spanning 8–20 nodes.
 
 ### Generation Process
@@ -182,6 +183,17 @@ visualize_graph(graph, "motif_example.png")
 The script saves a PNG with motif nodes highlighted in orange.
 Use absolute imports (e.g., `from motif_qa.src.utils import visualize_graph`) if you install the project as a package.
 
+### Interactive Viewer (Streamlit)
+
+Launch the interactive graph viewer:
+
+```bash
+streamlit run viewer.py
+```
+
+The sidebar lets you pick the dataset source (local JSONL or Hugging Face Hub), subset, split, and sample index.
+You can switch between an interactive PyVis view and a static NetworkX view to intuitively understand graph structures.
+
 ### Regenerating the Dataset Locally
 
 ```bash

generator.py

@@ -66,11 +66,16 @@ def create_ba_shapes_graphs(num_pos_samples: int, num_neg_samples: int):
 
 def create_tree_graphs(num_pos_samples: int, num_neg_samples: int, motif: Literal["cycle", "grid"]):
     """Create base datasets of tree graphs with and without motifs."""
-    MIN_NODES = 8
+    MIN_NODES = 10
     MAX_NODES = 20
+    motif_nnodes = {"cycle": 6, "grid": 9}
     motif_gen_args = {"cycle": {"num_nodes": 6}, "grid": None}
+
     pos_ds = ExplainerDataset(
-        graph_generator=UniformPruferTreeGraph(range(MIN_NODES - 5, MAX_NODES - 5 + 1), undirected=True),
+        graph_generator=UniformPruferTreeGraph(
+            range(MIN_NODES - motif_nnodes[motif], MAX_NODES - motif_nnodes[motif] + 1),
+            undirected=True,
+        ),
         motif_generator=motif,
         motif_generator_kwargs=motif_gen_args[motif],
         num_motifs=1,
@@ -82,6 +87,7 @@ def create_tree_graphs(num_pos_samples: int, num_neg_samples: int, motif: Litera
         num_graphs=num_neg_samples,
         transform=ShuffleNodes(),
     )
+
     return pos_ds, neg_ds
 
 
@@ -154,7 +160,7 @@ def format_yes_no_answer(motif_nodes: Iterable[int], edges: list[list[int]]) ->
     return "Yes, there is a motif." if list(motif_nodes) else "No, there is no motif."
 
 
-def format_classification_answer(motif_nodes: Iterable[int], edges: list[list[int]]) -> str:
+def format_two_motifs_answer(motif_nodes: Iterable[int], edges: list[list[int]]) -> str:
     motif_edges = [edge for edge in edges if edge[0] in motif_nodes and edge[1] in motif_nodes]
     if len(motif_edges) == 6:
         return "Yes, there is a house motif."
@@ -213,7 +219,7 @@ def main():
                 num_neg_samples=args.num_neg_samples,
             )
         case "ba_two_motifs":
-            formatter = format_classification_answer
+            formatter = format_two_motifs_answer
             pos_ds, neg_ds = create_two_motif_graphs(
                 num_pos_samples=args.num_pos_samples,
                 num_neg_samples=args.num_neg_samples,

viewer.py

@@ -0,0 +1,283 @@
+from __future__ import annotations
+
+import json
+import random
+from pathlib import Path
+
+import matplotlib.pyplot as plt
+import networkx as nx
+import streamlit as st
+import torch
+from torch_geometric.data import Data
+from torch_geometric.utils import is_undirected, to_networkx
+
+DATA_DIR = Path("data")
+DEFAULT_SUBSETS = ("ba_shapes", "tree_cycle", "tree_grid", "ba_two_motifs")
+
+
+def _available_subsets() -> list[str]:
+    if DATA_DIR.is_dir():
+        subsets = sorted([p.name for p in DATA_DIR.iterdir() if p.is_dir()])
+        if subsets:
+            return subsets
+    return list(DEFAULT_SUBSETS)
+
+
+@st.cache_data(show_spinner=False)
+def _load_jsonl(path: Path) -> list[dict]:
+    if not path.is_file():
+        raise FileNotFoundError(f"Could not find dataset file at {path}.")
+    with path.open("r", encoding="utf-8") as f:
+        return [json.loads(line) for line in f if line.strip()]
+
+
+@st.cache_data(show_spinner=False)
+def _load_local_examples(subset: str, split: str) -> list[dict]:
+    dataset_path = DATA_DIR / subset / f"{split}.jsonl"
+    return _load_jsonl(dataset_path)
+
+
+@st.cache_data(show_spinner=False)
+def _load_hf_examples(dataset_id: str, subset: str, split: str) -> list[dict]:
+    try:
+        from datasets import load_dataset
+    except ImportError as exc:
+        raise RuntimeError("The datasets package is required for Hugging Face loading.") from exc
+    dataset = load_dataset(dataset_id, name=subset, split=split)
+    return [dataset[i] for i in range(len(dataset))]
+
+
+def _build_graph(example: dict) -> Data:
+    num_nodes = int(example["nnodes"])
+    edges: list[list[int]] = example.get("edges", [])
+    undirected_edges: list[tuple[int, int]] = []
+    for src, dst in edges:
+        undirected_edges.append((src, dst))
+        if src != dst:
+            undirected_edges.append((dst, src))
+
+    if undirected_edges:
+        edge_index = torch.tensor(undirected_edges, dtype=torch.long).t().contiguous()
+    else:
+        edge_index = torch.empty((2, 0), dtype=torch.long)
+
+    graph = Data(edge_index=edge_index, num_nodes=num_nodes)
+    node_mask = torch.zeros(num_nodes, dtype=torch.float32)
+    motif_nodes = example.get("motif_nodes", [])
+    if motif_nodes:
+        node_mask[motif_nodes] = 1.0
+    graph.node_mask = node_mask
+    return graph
+
+
+def _graph_html(graph: Data, layout_seed: int) -> str:
+    if not is_undirected(graph.edge_index):
+        raise ValueError("The generated graph is not undirected.")
+
+    try:
+        from pyvis.network import Network
+    except ImportError as exc:
+        raise RuntimeError("pyvis is required for interactive graph visualization.") from exc
+
+    node_mask = getattr(graph, "node_mask", None)
+    motif_nodes = set(node_mask.nonzero(as_tuple=True)[0].tolist())
+
+    net = Network(height="650px", width="100%", directed=False, bgcolor="#ffffff")
+    net.barnes_hut()
+    net.set_options(
+        f"""
+        {{
+          "layout": {{
+            "randomSeed": {layout_seed}
+          }},
+          "physics": {{
+            "barnesHut": {{
+              "gravitationalConstant": -20000,
+              "springLength": 120,
+              "springConstant": 0.04
+            }},
+            "minVelocity": 0.75
+          }},
+          "nodes": {{
+            "shape": "dot",
+            "font": {{
+              "size": 24,
+              "face": "arial"
+            }}
+          }},
+          "edges": {{
+            "color": {{
+              "inherit": true
+            }},
+            "smooth": false
+          }}
+        }}
+        """
+    )
+
+    for node_id in range(graph.num_nodes):
+        size = 18
+        if node_id in motif_nodes:
+            color = "#ff8c00"
+            font_color = "#111111"
+        else:
+            color = "#87ceeb"
+            font_color = "#111111"
+        label = str(node_id)
+        net.add_node(node_id, label=label, color=color, size=size, font={"color": font_color})
+
+    edge_index = graph.edge_index.t().tolist()
+    seen = set()
+    for src, dst in edge_index:
+        if src == dst:
+            continue
+        key = (min(src, dst), max(src, dst))
+        if key in seen:
+            continue
+        seen.add(key)
+        net.add_edge(src, dst)
+
+    return net.generate_html()
+
+
+def _plot_graph(graph: Data, layout_seed: int) -> plt.Figure:
+    if not is_undirected(graph.edge_index):
+        raise ValueError("The generated graph is not undirected.")
+
+    G = to_networkx(graph, to_undirected=True)
+    pos = nx.spring_layout(G, seed=layout_seed)
+
+    node_mask = getattr(graph, "node_mask", None)
+    motif_nodes = node_mask.nonzero(as_tuple=True)[0].tolist()
+    normal_nodes = [n for n in G.nodes if n not in motif_nodes]
+
+    fig, ax = plt.subplots(figsize=(6, 4.5))
+    nx.draw_networkx_nodes(G, pos, nodelist=normal_nodes, node_color="skyblue", node_size=120, ax=ax)
+    nx.draw_networkx_nodes(G, pos, nodelist=motif_nodes, node_color="orange", node_size=120, ax=ax)
+    nx.draw_networkx_edges(G, pos, alpha=0.3, ax=ax)
+
+    labels = {n: str(n) for n in G.nodes()}
+    if normal_nodes:
+        nx.draw_networkx_labels(
+            G,
+            pos,
+            labels={n: labels[n] for n in normal_nodes},
+            font_size=10,
+            font_color="black",
+            ax=ax,
+        )
+    if motif_nodes:
+        nx.draw_networkx_labels(
+            G,
+            pos,
+            labels={n: labels[n] for n in motif_nodes},
+            font_size=10,
+            font_color="black",
+            ax=ax,
+        )
+
+    ax.axis("off")
+    fig.tight_layout()
+    return fig
+
+
+def render_sidebar() -> dict | None:
+    subsets = _available_subsets()
+    with st.sidebar:
+        dataset_source = st.radio("Dataset", ["Local JSONL", "Hugging Face Hub"], index=0)
+        dataset_id = st.text_input("Dataset ID", "naos-ku/motif-qa", disabled=dataset_source != "Hugging Face Hub")
+        subset = st.selectbox("Subset", options=subsets, index=0)
+        split_choice = st.selectbox("Split", options=["train", "val", "test"], index=0)
+
+    if dataset_source == "Local JSONL":
+        split_name = "val" if split_choice == "val" else split_choice
+        try:
+            examples = _load_local_examples(subset, split_name)
+        except FileNotFoundError as exc:
+            st.error(str(exc))
+            return None
+    else:
+        split_name = "validation" if split_choice == "val" else split_choice
+        try:
+            examples = _load_hf_examples(dataset_id, subset, split_name)
+        except Exception as exc:
+            st.error(str(exc))
+            return None
+
+    if not examples:
+        st.warning("No examples found for the selected configuration.")
+        return None
+
+    st.sidebar.caption(f"Samples: {len(examples)}")
+
+    if "sample_idx" not in st.session_state:
+        st.session_state.sample_idx = 0
+
+    def _pick_random_sample() -> None:
+        st.session_state.sample_idx = random.randrange(len(examples))
+
+    sample_idx = st.sidebar.number_input(
+        "Sample idx",
+        min_value=0,
+        max_value=len(examples) - 1,
+        value=st.session_state.sample_idx,
+        step=1,
+        key="sample_idx",
+    )
+    st.sidebar.button("Pick randomly", on_click=_pick_random_sample)
+    st.sidebar.divider()
+    st.sidebar.markdown("### Visualization Settings")
+    viz_method = st.sidebar.radio("Visualization method", ["Interactive (PyVis)", "Static (NetworkX)"], index=0)
+    layout_seed = None
+    if viz_method == "Static (NetworkX)":
+        layout_seed = st.sidebar.number_input("Layout seed", min_value=0, value=42, step=1)
+
+    st.sidebar.divider()
+    st.sidebar.link_button("Open HF dataset", "https://huggingface.co/datasets/naos-ku/motif-qa", icon="🤗")
+
+    return {
+        "examples": examples,
+        "sample_idx": sample_idx,
+        "viz_method": viz_method,
+        "layout_seed": layout_seed,
+    }
+
+
+def render_content(state: dict) -> None:
+    example = state["examples"][state["sample_idx"]]
+    graph = _build_graph(example)
+
+    with st.expander("Sample details", expanded=False):
+        num_edges = example.get("nedges", len(example.get("edges", [])))
+        st.write(f"Nodes: {example.get('nnodes', 0)}")
+        st.write(f"Edges: {num_edges}")
+        motif_nodes = example.get("motif_nodes", [])
+        st.write(f"Motif nodes: {motif_nodes if motif_nodes else 'None'}")
+        if "prompt" in example:
+            st.markdown("**Prompt**")
+            st.write(example["prompt"])
+        if "response" in example:
+            st.markdown("**Response**")
+            st.write(example["response"])
+
+    if state["viz_method"] == "Static (NetworkX)":
+        fig = _plot_graph(graph, layout_seed=state["layout_seed"] or 42)
+        st.pyplot(fig, clear_figure=True)
+        plt.close(fig)
+    else:
+        html = _graph_html(graph, layout_seed=42)
+        st.components.v1.html(html, height=700, scrolling=True)
+
+
+def main() -> None:
+    st.set_page_config(page_title="MotifQA Graph Viewer", layout="wide")
+    st.title("MotifQA Graph Viewer")
+
+    state = render_sidebar()
+    if state is None:
+        return
+    render_content(state)
+
+
+if __name__ == "__main__":
+    main()