Download src/components/data_loader.py from jmisidro/HSeg_Demo: direct link, hf CLI and curl.
- Browser
- Download file 4.73 kB
-
https://huggingface.co/spaces/jmisidro/HSeg_Demo/resolve/main/src/components/data_loader.py
- Command line
-
hf download hf://spaces/jmisidro/HSeg_Demo/src/components/data_loader.py
-
curl -L -o data_loader.py https://huggingface.co/spaces/jmisidro/HSeg_Demo/resolve/main/src/components/data_loader.py
4.73 kB
| """ | |
| Data Loader for LLM Subject Extraction Demo | |
| Handles loading of LLM evaluation run results from the results/llm_evaluation directory. | |
| """ | |
| import json | |
| import re | |
| from pathlib import Path | |
| from typing import Dict, List, Any, Optional | |
| # --------------------------------------------------------------------------- | |
| # Colour palette for topic labels | |
| # --------------------------------------------------------------------------- | |
| TOPIC_COLORS = [ | |
| "#4C72B0", "#DD8452", "#55A868", "#C44E52", "#8172B2", | |
| "#937860", "#DA8BC3", "#8C8C8C", "#CCB974", "#64B5CD", | |
| "#3B7ABF", "#E07B39", "#3D8F5C", "#A63030", "#6B5FA6", | |
| "#7A6550", "#C66FAD", "#6E6E6E", "#B39C5A", "#4EA3B8", | |
| ] | |
| class LLMEvalDataLoader: | |
| """Loads and caches LLM evaluation run data from disk.""" | |
| def __init__(self, results_root: str): | |
| self._run_cache: Dict[str, Dict[str, Any]] = {} | |
| example_data_path = Path(__file__).parent.parent / "example_data.json" | |
| if example_data_path.exists(): | |
| try: | |
| self._run_cache = json.loads(example_data_path.read_text(encoding="utf-8")) | |
| except Exception as e: | |
| print(f"Error loading example data: {e}") | |
| self._topic_color_map: Dict[str, str] = {} | |
| self._color_idx = 0 | |
| # ------------------------------------------------------------------ | |
| # Run discovery | |
| # ------------------------------------------------------------------ | |
| def get_available_runs(self) -> List[Dict[str, str]]: | |
| """Return metadata for every evaluation run found in the results root.""" | |
| runs: List[Dict[str, str]] = [] | |
| for run_id, run in self._run_cache.items(): | |
| config = run.get("config", {}) | |
| meta = { | |
| "run_id": run_id, | |
| "path": "", | |
| "backend": config.get("pipeline_config", {}).get("backend", "unknown"), | |
| "model_name": config.get("pipeline_config", {}).get("model_name", "unknown"), | |
| "timestamp": config.get("timestamp", ""), | |
| "split": config.get("split", ""), | |
| } | |
| runs.append(meta) | |
| return runs | |
| # ------------------------------------------------------------------ | |
| # Per-run data | |
| # ------------------------------------------------------------------ | |
| def load_run(self, run_id: str) -> Optional[Dict[str, Any]]: | |
| """Load the full data for a single evaluation run.""" | |
| return self._run_cache.get(run_id) | |
| # ------------------------------------------------------------------ | |
| # Helper accessors | |
| # ------------------------------------------------------------------ | |
| def get_document(self, run_id: str, doc_id: str) -> Optional[Dict[str, Any]]: | |
| run = self.load_run(run_id) | |
| if run is None: | |
| return None | |
| return run.get("documents", {}).get(doc_id) | |
| def get_aggregate(self, run_id: str) -> Optional[Dict[str, Any]]: | |
| run = self.load_run(run_id) | |
| if run is None: | |
| return None | |
| return run.get("aggregate") | |
| def get_config(self, run_id: str) -> Optional[Dict[str, Any]]: | |
| run = self.load_run(run_id) | |
| if run is None: | |
| return None | |
| return run.get("config") | |
| # ------------------------------------------------------------------ | |
| # Topic colour management | |
| # ------------------------------------------------------------------ | |
| def get_topic_color(self, topic: str) -> str: | |
| if topic not in self._topic_color_map: | |
| self._topic_color_map[topic] = TOPIC_COLORS[ | |
| self._color_idx % len(TOPIC_COLORS) | |
| ] | |
| self._color_idx += 1 | |
| return self._topic_color_map[topic] | |
| def get_all_topics_in_run(self, run_id: str) -> List[str]: | |
| """Return the sorted list of all unique GT topics seen in a run.""" | |
| run = self.load_run(run_id) | |
| if not run: | |
| return [] | |
| topics: set = set() | |
| for doc_data in run.get("documents", {}).values(): | |
| for alignment in doc_data.get("aligned_comparison", {}).get("subjects_alignment", []): | |
| gt = alignment.get("matched_gt", {}) | |
| topics.update(gt.get("topics", [])) | |
| return sorted(topics) | |
| # ------------------------------------------------------------------ | |
| # Municipality extraction helper | |
| # ------------------------------------------------------------------ | |
| def parse_municipality(doc_id: str) -> str: | |
| return doc_id.split("_")[0] if "_" in doc_id else doc_id | |
| def parse_date(doc_id: str) -> str: | |
| m = re.search(r"(\d{4}-\d{2}-\d{2})", doc_id) | |
| return m.group(1) if m else "" | |