File size: 5,772 Bytes
b59b737 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 | import hashlib
from flask import Blueprint, request, jsonify
from datasets import load_dataset
bp = Blueprint("adaevolve_datasets", __name__, url_prefix="/api/adaevolve/datasets")
_cache: dict[str, dict] = {}
def _make_id(repo: str, split: str) -> str:
key = f"{repo}:{split}"
return hashlib.md5(key.encode()).hexdigest()[:12]
def _build_iteration_summary(row: dict, idx: int) -> dict:
"""Build a summary for one iteration row."""
return {
"index": idx,
"iteration": row.get("iteration", idx),
"island_id": row.get("island_id", 0),
"score": row.get("score", 0.0),
"best_score": row.get("best_score", 0.0),
"delta": row.get("delta", 0.0),
"adaptation_type": row.get("adaptation_type", ""),
"exploration_intensity": row.get("exploration_intensity", 0.0),
"is_valid": row.get("is_valid", False),
"task_id": row.get("task_id", ""),
"meta_guidance_tactic": row.get("meta_guidance_tactic", ""),
"tactic_approach_type": row.get("tactic_approach_type", ""),
}
def _build_summary_stats(iterations: list[dict]) -> dict:
"""Build aggregate stats across all iterations."""
adaptation_counts: dict[str, int] = {}
island_best_scores: dict[int, float] = {}
global_best = 0.0
for it in iterations:
atype = it.get("adaptation_type", "unknown")
adaptation_counts[atype] = adaptation_counts.get(atype, 0) + 1
iid = it.get("island_id", 0)
score = it.get("best_score", 0.0)
if iid not in island_best_scores or score > island_best_scores[iid]:
island_best_scores[iid] = score
if score > global_best:
global_best = score
return {
"adaptation_counts": adaptation_counts,
"island_best_scores": {str(k): v for k, v in island_best_scores.items()},
"global_best": global_best,
"n_islands": len(island_best_scores),
}
@bp.route("/load", methods=["POST"])
def load_dataset_endpoint():
data = request.get_json()
repo = data.get("repo", "").strip()
if not repo:
return jsonify({"error": "repo is required"}), 400
split = data.get("split", "train")
try:
ds = load_dataset(repo, split=split)
except Exception as e:
return jsonify({"error": f"Failed to load dataset: {e}"}), 400
ds_id = _make_id(repo, split)
# Build iteration summaries
iterations = []
for i in range(len(ds)):
row = ds[i]
summary = _build_iteration_summary(row, i)
iterations.append(summary)
summary_stats = _build_summary_stats(iterations)
_cache[ds_id] = {
"repo": repo,
"split": split,
"dataset": ds,
"iterations": iterations,
}
short_name = repo.rsplit("/", 1)[-1] if "/" in repo else repo
return jsonify({
"id": ds_id,
"repo": repo,
"name": short_name,
"split": split,
"iterations": iterations,
"n_iterations": len(iterations),
"summary_stats": summary_stats,
})
@bp.route("/", methods=["GET"])
def list_datasets():
result = []
for ds_id, info in _cache.items():
iterations = info["iterations"]
summary_stats = _build_summary_stats(iterations)
result.append({
"id": ds_id,
"repo": info["repo"],
"name": info["repo"].rsplit("/", 1)[-1] if "/" in info["repo"] else info["repo"],
"split": info["split"],
"n_iterations": len(iterations),
"iterations": iterations,
"summary_stats": summary_stats,
})
return jsonify(result)
@bp.route("/<ds_id>/iterations", methods=["GET"])
def get_iterations(ds_id):
if ds_id not in _cache:
return jsonify({"error": "Dataset not loaded"}), 404
return jsonify(_cache[ds_id]["iterations"])
@bp.route("/<ds_id>/iteration/<int:idx>", methods=["GET"])
def get_iteration(ds_id, idx):
"""Get full detail for one iteration including prompt, reasoning, code."""
if ds_id not in _cache:
return jsonify({"error": "Dataset not loaded"}), 404
info = _cache[ds_id]
if idx < 0 or idx >= len(info["dataset"]):
return jsonify({"error": f"Iteration index {idx} out of range"}), 404
row = info["dataset"][idx]
return jsonify({
"index": idx,
"iteration": row.get("iteration", idx),
"island_id": row.get("island_id", 0),
"score": row.get("score", 0.0),
"best_score": row.get("best_score", 0.0),
"delta": row.get("delta", 0.0),
"adaptation_type": row.get("adaptation_type", ""),
"exploration_intensity": row.get("exploration_intensity", 0.0),
"is_valid": row.get("is_valid", False),
"task_id": row.get("task_id", ""),
"prompt_text": row.get("prompt_text", ""),
"reasoning_trace": row.get("reasoning_trace", ""),
"program_code": row.get("program_code", ""),
"meta_guidance_tactic": row.get("meta_guidance_tactic", ""),
"tactic_approach_type": row.get("tactic_approach_type", ""),
})
@bp.route("/available-runs", methods=["GET"])
def available_runs():
"""Query PROJECT-MANIFEST for adaevolve-compatible datasets."""
try:
from backend.api.manifest import query_runs
# Use empty prefix to get ALL datasets (not just adaevolve-prefixed ones)
# Users want to browse any dataset, not just ones with a specific prefix
return jsonify(query_runs(""))
except Exception as e:
return jsonify({"error": f"Failed to query manifest: {e}"}), 500
@bp.route("/<ds_id>", methods=["DELETE"])
def unload_dataset(ds_id):
if ds_id in _cache:
del _cache[ds_id]
return jsonify({"status": "ok"})
|