File size: 3,750 Bytes
165da3c |
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 |
# Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
# SPDX-License-Identifier: CC-BY-NC-4.0
import json
import random
from pathlib import Path
from typing import Dict, List
from datetime import datetime
from loguru import logger
from models import DocumentAsset
class SplitManager:
"""Manages train/test/validation splits for documents."""
def __init__(
self,
train_ratio: float = 0.6,
test_ratio: float = 0.25,
validation_ratio: float = 0.15,
random_seed: int = 42
):
self.train_ratio = train_ratio
self.test_ratio = test_ratio
self.validation_ratio = validation_ratio
self.random_seed = random_seed
if abs(train_ratio + test_ratio + validation_ratio - 1.0) > 0.001:
raise ValueError("Split ratios must sum to 1.0")
def create_split(
self,
documents_by_type: Dict[str, List[DocumentAsset]]
) -> Dict[str, Dict[str, List[str]]]:
"""Create stratified train/test/validation split.
Returns:
Dict with structure: {split_name: {doc_type: [doc_names]}}
"""
random.seed(self.random_seed)
splits = {
'train': {},
'test': {},
'validation': {}
}
for doc_type, documents in documents_by_type.items():
doc_names = [doc.doc_name for doc in documents]
random.shuffle(doc_names)
total = len(doc_names)
train_end = int(total * self.train_ratio)
test_end = train_end + int(total * self.test_ratio)
splits['train'][doc_type] = doc_names[:train_end]
splits['test'][doc_type] = doc_names[train_end:test_end]
splits['validation'][doc_type] = doc_names[test_end:]
logger.info(f"Created split: train={self._count_docs(splits['train'])}, "
f"test={self._count_docs(splits['test'])}, "
f"val={self._count_docs(splits['validation'])}")
return splits
def save_split(self, splits: Dict, output_path: str):
"""Save split mapping to JSON."""
Path(output_path).parent.mkdir(parents=True, exist_ok=True)
split_data = {
'created_at': datetime.now().isoformat(),
'split_config': {
'train_ratio': self.train_ratio,
'test_ratio': self.test_ratio,
'validation_ratio': self.validation_ratio,
'random_seed': self.random_seed,
'stratified': True
},
'splits': splits,
'statistics': {
'train': self._get_statistics(splits['train']),
'test': self._get_statistics(splits['test']),
'validation': self._get_statistics(splits['validation'])
}
}
with open(output_path, 'w') as f:
json.dump(split_data, f, indent=2)
logger.info(f"Saved split mapping to {output_path}")
def load_split(self, split_path: str) -> Dict[str, Dict[str, List[str]]]:
"""Load split mapping from JSON."""
with open(split_path, 'r') as f:
split_data = json.load(f)
logger.info(f"Loaded split mapping from {split_path}")
return split_data['splits']
def _count_docs(self, split: Dict[str, List[str]]) -> int:
"""Count total documents in a split."""
return sum(len(docs) for docs in split.values())
def _get_statistics(self, split: Dict[str, List[str]]) -> Dict[str, int]:
"""Get statistics for a split."""
stats = {'total': self._count_docs(split)}
for doc_type, docs in split.items():
stats[doc_type] = len(docs)
return stats
|