File size: 7,331 Bytes
5fed0fc |
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 |
"""
Pair expansion and management for batch evaluation.
A Pair represents a (solution, problem) combination to evaluate.
"""
import hashlib
from dataclasses import dataclass
from pathlib import Path
from typing import List, Optional
from ..models import get_model_prefix, sanitize_problem_name
@dataclass
class Pair:
"""Represents a solution-problem pair for evaluation."""
solution: str # Solution identifier (e.g., "gpt5_flash_attn")
problem: str # Problem identifier (e.g., "flash_attn")
@property
def id(self) -> str:
"""Unique identifier for this pair."""
return f"{self.solution}:{self.problem}"
@property
def safe_name(self) -> str:
"""Filesystem-safe name for this pair."""
base = f"{self.solution}-{self.problem}"
digest = hashlib.md5(base.encode("utf-8")).hexdigest()[:8]
sanitized = _sanitize_name(base)
suffix = f"-{digest}"
max_len = 63
available = max_len - len(suffix)
trimmed = sanitized[:available].rstrip("-")
return _sanitize_name(f"{trimmed}{suffix}")
def __hash__(self):
return hash(self.id)
def __eq__(self, other):
if isinstance(other, Pair):
return self.id == other.id
return False
def _sanitize_name(name: str) -> str:
"""Sanitize a name to be a valid cluster/file name."""
cleaned = []
valid = "abcdefghijklmnopqrstuvwxyz0123456789-"
last_dash = False
for ch in name.lower():
if ch in valid:
cleaned.append(ch)
last_dash = ch == "-"
else:
if not last_dash:
cleaned.append("-")
last_dash = True
sanitized = "".join(cleaned).strip("-")
return sanitized or "job"
# get_model_prefix and sanitize_problem_name imported from ..models
def expand_pairs(
problems: List[str],
models: List[str],
variants: Optional[List[int]] = None,
*,
solutions_dir: Optional[Path] = None,
validate_paths: bool = True,
) -> List[Pair]:
"""
Expand problems × models × variants into pairs.
Args:
problems: List of problem IDs (e.g., ["flash_attn", "cross_entropy"])
models: List of model names (e.g., ["gpt-5", "claude-sonnet-4-5"])
variants: List of variant indices (default: [0] for no suffix)
solutions_dir: Directory containing solutions (for validation)
validate_paths: Whether to validate solution paths exist
Returns:
List of Pair objects
"""
if variants is None:
variants = [0]
pairs: List[Pair] = []
for problem in problems:
problem_name = sanitize_problem_name(problem)
for model in models:
model_prefix = get_model_prefix(model)
for variant_idx in variants:
suffix = "" if variant_idx == 0 else f"_{variant_idx}"
solution_name = f"{model_prefix}_{problem_name}{suffix}"
if validate_paths and solutions_dir:
solution_path = solutions_dir / solution_name
if not solution_path.exists():
continue
pairs.append(Pair(solution=solution_name, problem=problem))
return pairs
def read_pairs_file(path: Path) -> List[Pair]:
"""
Read pairs from a pairs file.
Format: one pair per line as "solution:problem"
Lines starting with # are comments.
"""
pairs: List[Pair] = []
with path.open("r", encoding="utf-8") as f:
for line in f:
stripped = line.strip()
if not stripped or stripped.startswith("#"):
continue
if ":" not in stripped:
raise ValueError(f"Invalid pair line (expected solution:problem): {stripped}")
solution, problem = stripped.split(":", 1)
pairs.append(Pair(solution=solution.strip(), problem=problem.strip()))
return pairs
def read_problems_file(path: Path) -> List[str]:
"""Read problems from a problems file (one per line)."""
problems: List[str] = []
with path.open("r", encoding="utf-8") as f:
for line in f:
stripped = line.strip()
if not stripped or stripped.startswith("#"):
continue
# Normalize: remove 'research/' prefix if present
if stripped.startswith("research/"):
stripped = stripped[len("research/"):]
problems.append(stripped)
return problems
def read_models_file(path: Path) -> List[str]:
"""Read models from a models file (one per line)."""
models: List[str] = []
with path.open("r", encoding="utf-8") as f:
for line in f:
stripped = line.strip()
if not stripped or stripped.startswith("#"):
continue
models.append(stripped)
return models
def read_variants_file(path: Path) -> List[int]:
"""Read variant indices from a file (one per line)."""
variants: List[int] = []
if not path.exists():
return [0] # Default: just index 0 (no suffix)
with path.open("r", encoding="utf-8") as f:
for line in f:
stripped = line.strip()
if not stripped or stripped.startswith("#"):
continue
try:
variants.append(int(stripped))
except ValueError:
pass
return variants if variants else [0]
def read_solution_config(solution_dir: Path) -> Optional[str]:
"""
Read problem from a solution's config.yaml.
Returns:
Problem path (e.g., "flash_attn") or None if not found.
"""
config_file = solution_dir / "config.yaml"
if not config_file.exists():
return None
try:
import yaml
with config_file.open("r", encoding="utf-8") as f:
config = yaml.safe_load(f)
if config and isinstance(config, dict):
return config.get("problem")
except Exception:
# Fallback: simple parsing for "problem: value"
try:
content = config_file.read_text(encoding="utf-8")
for line in content.splitlines():
line = line.strip()
if line.startswith("problem:"):
return line.split(":", 1)[1].strip()
except Exception:
pass
return None
def scan_solutions_dir(solutions_dir: Path) -> List[Pair]:
"""
Scan solutions directory and build pairs from config.yaml files.
Args:
solutions_dir: Path to solutions directory
Returns:
List of Pair objects for solutions that have valid config.yaml
"""
pairs: List[Pair] = []
if not solutions_dir.is_dir():
return pairs
for solution_path in sorted(solutions_dir.iterdir()):
if not solution_path.is_dir() or solution_path.name.startswith("."):
continue
problem = read_solution_config(solution_path)
if problem:
pairs.append(Pair(solution=solution_path.name, problem=problem))
return pairs
|