File size: 11,628 Bytes
07f29dc | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 | """
Quick single-instance cascade validation.
Runs the cascade agent + verification on ONE instance.
This is the minimal proof that the cascade works end-to-end.
Usage:
python quick_validate.py
python quick_validate.py --instance django__django-11815
"""
import json
import os
import re
import subprocess
import sys
import tempfile
import time
import traceback
from datetime import datetime
from pathlib import Path
from typing import Optional, Tuple
from huggingface_hub import InferenceClient
# ============================================================
# Pick the easiest instance first
# ============================================================
DEFAULT_INSTANCE = "django__django-14315" # django bug with clean fix
# For T1/T2 models (free HF inference)
T1_MODEL = "meta-llama/Llama-3.1-8B-Instruct"
T2_MODEL = "meta-llama/Llama-3.3-70B-Instruct"
def run(cmd, cwd=None, timeout=120):
result = subprocess.run(cmd, cwd=cwd, capture_output=True, text=True, timeout=timeout, shell=True)
return result.returncode, result.stdout, result.stderr
def call_model(client, messages, max_tokens=4096):
"""Call HF inference, return (text, input_tokens, output_tokens)."""
try:
completion = client.chat.completions.create(
model=client.model,
messages=messages,
max_tokens=max_tokens,
temperature=0.2,
)
text = completion.choices[0].message.content
itok = completion.usage.prompt_tokens if hasattr(completion, 'usage') and completion.usage else 0
otok = completion.usage.completion_tokens if hasattr(completion, 'usage') and completion.usage else 0
return text, itok, otok
except Exception as e:
return f"[ERROR: {e}]", 0, 0
def extract_patch(text):
"""Extract a diff/patch from model output."""
for tag in ['patch', 'diff']:
m = re.search(rf'<{tag}>(.*?)</{tag}>', text, re.DOTALL)
if m:
return m.group(1).strip()
for block in ['diff', 'patch']:
m = re.search(rf'```{block}\s*\n(.*?)```', text, re.DOTALL)
if m:
return m.group(1).strip()
diff_match = re.search(r'(diff --git a/.*?(?:\n(?:@@|\+\+\+|diff --git|```|</).*)*)', text, re.DOTALL)
if diff_match:
return diff_match.group(1).strip()
return None
def run_cascade(instance, repo_dir):
"""Run T1 then T2. Returns {patch, tier, turns, tokens}."""
problem = instance.get("problem_statement", "")
system = f"""You are fixing a bug in {instance['repo']}. Repository is at {repo_dir}.
Output format:
- Bash commands: <bash>command</bash>
- Final patch: <patch>your diff here</patch>
- Done: <submit>Done</submit>
First explore the codebase to understand the issue, then make a minimal fix and verify it."""
messages = [
{"role": "system", "content": system},
{"role": "user", "content": f"PROBLEM:\n{problem}\n\nStart by exploring the repository."}
]
tiers = [
("T1", T1_MODEL, 30),
("T2", T2_MODEL, 30),
]
for tier_name, model_id, max_turns in tiers:
print(f"\n[{tier_name}] Running {model_id}...")
client = InferenceClient(model_id)
tier_turns = 0
tier_itok = 0
tier_otok = 0
for turn in range(max_turns):
text, itok, otok = call_model(client, messages, max_tokens=4096)
tier_turns += 1
tier_itok += itok
tier_otok += otok
messages.append({"role": "assistant", "content": text})
# Extract patch
patch = extract_patch(text)
if patch:
print(f" ✅ Patch found ({len(patch)} chars) at turn {turn+1}")
return {"patch": patch, "tier": tier_name, "turns": tier_turns, "input_tokens": tier_itok, "output_tokens": tier_otok}
# Execute bash commands
cmds = re.findall(r'<bash>(.*?)</bash>', text, re.DOTALL)
for cmd in cmds:
cmd = cmd.strip()
print(f" $ {cmd[:80]}...")
rc, stdout, stderr = run(cmd, cwd=str(repo_dir), timeout=30)
output = (stdout + stderr)[:1500]
if rc != 0:
output += f" [EXIT:{rc}]"
messages.append({"role": "user", "content": f"<output>\n{output}\n</output>"})
if "<submit>" in text:
break
return {"patch": None, "tier": None, "turns": 0, "input_tokens": 0, "output_tokens": 0}
def verify_patch(instance, model_patch, repo_dir, env_name=None):
"""Apply patch + test_patch, run FAIL_TO_PASS tests."""
base_commit = instance.get("base_commit", "")
test_patch = instance.get("test_patch", "")
f2p = instance.get("FAIL_TO_PASS", [])
if not base_commit or not test_patch or not f2p:
return {"resolved": False, "error": "missing base_commit/test_patch/FAIL_TO_PASS"}
# Reset
run(f"git checkout -f {base_commit}", cwd=str(repo_dir))
# Apply model patch
patch_file = repo_dir / "_aco.patch"
patch_file.write_text(model_patch)
rc, out, err = run(f"git apply --check {patch_file}", cwd=str(repo_dir))
if rc != 0:
return {"resolved": False, "error": f"patch --check: {err[:200]}"}
rc, out, err = run(f"git apply {patch_file}", cwd=str(repo_dir))
if rc != 0:
rc, out, err = run(f"git apply --reject {patch_file}", cwd=str(repo_dir))
if rc != 0:
return {"resolved": False, "error": f"patch apply: {err[:200]}"}
# Apply test patch
test_file = repo_dir / "_aco_test.patch"
test_file.write_text(test_patch)
rc, out, err = run(f"git apply --check {test_file}", cwd=str(repo_dir))
if rc == 0:
run(f"git apply {test_file}", cwd=str(repo_dir))
# Run FAIL_TO_PASS
cmd_prefix = f"conda run -n {env_name} " if env_name else ""
cmd = f"{cmd_prefix}python -m pytest -v --tb=short -x {' '.join(f2p)}"
print(f" Running: pytest {' '.join(f2p[:2])}...")
rc, out, err = run(cmd, cwd=str(repo_dir), timeout=300)
if rc == 0:
# Check regressions
p2p = instance.get("PASS_TO_PASS", [])
if p2p:
cmd2 = f"{cmd_prefix}python -m pytest -v --tb=short -x {' '.join(p2p[:15])}"
rc2, out2, err2 = run(cmd2, cwd=str(repo_dir), timeout=300)
if rc2 == 0:
return {"resolved": True, "regressions": False}
else:
return {"resolved": False, "error": f"regression: {(out2+err2)[:200]}", "regressions": True}
return {"resolved": True, "regressions": False}
# Count failures
failures = [l.strip() for l in (out+err).split('\n') if 'FAILED' in l]
return {"resolved": False, "error": f"{len(failures)} F2P failures", "failures": failures[:5]}
def main():
from datasets import load_dataset
instance_id = sys.argv[1] if len(sys.argv) > 1 else DEFAULT_INSTANCE
print(f"Validating: {instance_id}")
ds = load_dataset("princeton-nlp/SWE-bench_Verified", split="test")
instance = None
for row in ds:
if row["instance_id"] == instance_id:
instance = dict(row)
break
if not instance:
print(f"Instance {instance_id} not found!")
sys.exit(1)
print(f" Repo: {instance['repo']}")
print(f" Base: {instance['base_commit'][:12]}")
print(f" F2P: {len(instance.get('FAIL_TO_PASS', []))} tests")
with tempfile.TemporaryDirectory(prefix=f"aco_quick_") as tmpdir:
repo_dir = Path(tmpdir) / "repo"
env_name = f"aco_q_{instance_id.replace('__','_').replace('-','_')[:30]}"
# Clone
repo = instance["repo"]
url = f"https://github.com/{repo}.git"
print(f"\n[CLONE] {url}")
rc, out, err = run(f"git clone --depth 50 {url} {repo_dir}", timeout=180)
if rc != 0:
rc, out, err = run(f"git clone {url} {repo_dir}", timeout=600)
if rc != 0:
print(f"CLONE FAILED: {err[:300]}")
sys.exit(1)
# Set up conda env
env_commit = instance.get("environment_setup_commit", "")
if env_commit:
run(f"cd {repo_dir} && git fetch origin {env_commit}", timeout=60)
run(f"cd {repo_dir} && git checkout {env_commit}", timeout=30)
env_yml = None
for c in ["environment.yml", "dev/environment.yml", ".github/environment.yml"]:
if (repo_dir / c).exists():
env_yml = c
break
print(f"\n[ENV] Creating conda env '{env_name}'...")
if env_yml:
rc, out, err = run(f"cd {repo_dir} && conda env create -f {env_yml} -n {env_name} --quiet", timeout=600)
else:
rc, out, err = run(f"conda create -n {env_name} python=3.10 pip -y --quiet", timeout=300)
if rc != 0:
print(f"ENV SETUP FAILED: {err[:300]}")
sys.exit(1)
# Install repo at base_commit
base_commit = instance["base_commit"]
run(f"cd {repo_dir} && git fetch origin {base_commit}", timeout=60)
run(f"cd {repo_dir} && git checkout {base_commit}", timeout=30)
rc, out, err = run(f"cd {repo_dir} && conda run -n {env_name} pip install -e . --quiet", timeout=300)
if rc != 0:
print(f"PIP INSTALL FAILED (continuing): {err[:200]}")
print(f"\n[CASCADE] Running agent...")
t0 = time.time()
agent_result = run_cascade(instance, repo_dir)
agent_time = time.time() - t0
print(f" Patch: {'FOUND' if agent_result['patch'] else 'NOT FOUND'}")
print(f" Tier: {agent_result['tier']}")
print(f" Time: {agent_time:.1f}s")
if not agent_result["patch"]:
print("FAILED: No patch produced")
sys.exit(1)
print(f"\n[VERIFY] Testing patch...")
verify_result = verify_patch(instance, agent_result["patch"], repo_dir, env_name)
print(f"\n{'='*60}")
print(f"RESULT: {'✅ RESOLVED' if verify_result['resolved'] else '❌ NOT RESOLVED'}")
print(f"{'='*60}")
print(f" Instance: {instance_id}")
print(f" Tier: {agent_result['tier']}")
print(f" Turns: {agent_result['turns']}")
print(f" Tokens: {agent_result['input_tokens']} in / {agent_result['output_tokens']} out")
print(f" Agent time: {agent_time:.1f}s")
if not verify_result["resolved"]:
print(f" Error: {verify_result.get('error', 'unknown')}")
# Save result
final = {
"instance_id": instance_id,
"repo": instance["repo"],
"timestamp": datetime.now().isoformat(),
"resolved": verify_result["resolved"],
"tier": agent_result["tier"],
"turns": agent_result["turns"],
"input_tokens": agent_result["input_tokens"],
"output_tokens": agent_result["output_tokens"],
"agent_time_seconds": agent_time,
"error": verify_result.get("error"),
}
result_path = f"quick_validate_{instance_id}.json"
with open(result_path, "w") as f:
json.dump(final, f, indent=2)
print(f"\n Saved: {result_path}")
# Cleanup
run(f"conda env remove -n {env_name} -y --quiet", timeout=30)
return 0 if verify_result["resolved"] else 1
if __name__ == "__main__":
sys.exit(main())
|