File size: 11,074 Bytes
e425a8a d2ab04b e425a8a d2ab04b e425a8a d2ab04b e425a8a d2ab04b e425a8a d2ab04b e425a8a d2ab04b e425a8a d2ab04b e425a8a d2ab04b e425a8a d2ab04b e425a8a d2ab04b e425a8a d2ab04b e425a8a e5ab9ff e425a8a e5ab9ff e425a8a e5ab9ff e425a8a e5ab9ff e425a8a e5ab9ff e425a8a e5ab9ff e425a8a d2ab04b e5ab9ff fdaabfa e5ab9ff d2ab04b e425a8a d2ab04b e425a8a d2ab04b e425a8a e5ab9ff e425a8a e5ab9ff e425a8a e5ab9ff e425a8a e5ab9ff e425a8a e5ab9ff e425a8a e5ab9ff e425a8a d2ab04b fdaabfa e5ab9ff d2ab04b e425a8a d2ab04b e425a8a d2ab04b e425a8a | 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 | """
Shared utilities for polarity and topic scoring pipelines.
Provides common functions for model loading, prediction, and result saving.
"""
import re
import torch
import pandas as pd
from pathlib import Path
from tqdm import tqdm
from transformers import AutoTokenizer, AutoModelForSequenceClassification
def find_available_years(data_dir: Path) -> list:
"""
Auto-detect years by scanning data directory for all_reviews_*.csv files.
Args:
data_dir: Path to directory containing processed review data
Returns:
Sorted list of years found
"""
years = []
if data_dir.exists():
for file in data_dir.glob("all_reviews_*.csv"):
match = re.search(r'all_reviews_(\d{4})\.csv', file.name)
if match:
years.append(int(match.group(1)))
return sorted(years)
def _local_model_available(model_dir: Path) -> bool:
"""Check if a local model directory has the required files."""
if not model_dir.exists():
return False
# Accept either pytorch_model.bin or safetensors
has_weights = (model_dir / "pytorch_model.bin").exists() or (model_dir / "model.safetensors").exists()
return has_weights and (model_dir / "config.json").exists()
def load_model_and_tokenizer(model_dir: Path, device: str = "cuda", hub_fallback: str = None):
"""
Load a model and tokenizer from a local directory, or fall back to HuggingFace Hub.
Args:
model_dir: Path to local model directory
device: Device to load model onto ("cuda" or "cpu")
hub_fallback: HuggingFace Hub model ID to use if local files are missing
Returns:
Tuple of (tokenizer, model, device_obj)
"""
model_source = str(model_dir)
if not _local_model_available(model_dir):
if hub_fallback:
print(f" Local model not found at {model_dir}")
print(f" Falling back to HuggingFace Hub: {hub_fallback}")
model_source = hub_fallback
else:
raise FileNotFoundError(f"Model not found at {model_dir} and no hub fallback configured")
try:
tokenizer = AutoTokenizer.from_pretrained(model_source)
model = AutoModelForSequenceClassification.from_pretrained(model_source)
model.eval()
# Move to device
device_obj = torch.device(device if torch.cuda.is_available() else "cpu")
model.to(device_obj)
return tokenizer, model, device_obj
except Exception as e:
raise RuntimeError(f"Failed to load model from {model_source}: {e}")
def predict_batch(sentences: list, tokenizer, model, device, max_length: int = 512) -> list:
"""
Run batch predictions on a list of sentences.
Args:
sentences: List of sentence strings to predict
tokenizer: Tokenizer instance
model: Model instance
device: Device object for computation
max_length: Maximum token length (default: 512 for BERT-like models)
Returns:
List of predicted class IDs (integers)
"""
if not sentences:
return []
try:
inputs = tokenizer(
sentences,
return_tensors="pt",
padding=True,
truncation=True,
max_length=max_length
).to(device)
with torch.no_grad():
outputs = model(**inputs)
predictions = torch.argmax(outputs.logits, dim=1).cpu().tolist()
return predictions
except Exception as e:
raise RuntimeError(f"Prediction failed: {e}")
def save_polarity_results(output_path: Path, results: list) -> None:
"""
Save polarity scoring results to CSV.
Expected result format:
[
{"id": review_id, "sentence": sentence_text, "score": float, "label": int},
...
]
Args:
output_path: Path to output CSV file
results: List of result dictionaries
"""
output_path.parent.mkdir(parents=True, exist_ok=True)
df = pd.DataFrame(results)
df.to_csv(output_path, index=False)
def save_topic_results(output_path: Path, results: list) -> None:
"""
Save topic scoring results to CSV.
Expected result format:
[
{"id": review_id, "sentence": sentence_text, "topic_id": int, "topic_label": str},
...
]
Args:
output_path: Path to output CSV file
results: List of result dictionaries
"""
output_path.parent.mkdir(parents=True, exist_ok=True)
df = pd.DataFrame(results)
df.to_csv(output_path, index=False)
def validate_input_file(input_path: Path, required_columns: list) -> pd.DataFrame:
"""
Validate that input CSV file exists and has required columns.
Args:
input_path: Path to CSV file
required_columns: List of column names that must exist
Returns:
Loaded DataFrame
Raises:
FileNotFoundError: If file doesn't exist
ValueError: If required columns are missing
"""
if not input_path.exists():
raise FileNotFoundError(f"Input file not found: {input_path}")
try:
df = pd.read_csv(input_path)
except Exception as e:
raise ValueError(f"Failed to read CSV {input_path}: {e}")
missing_cols = set(required_columns) - set(df.columns)
if missing_cols:
raise ValueError(f"Missing required columns: {missing_cols}")
return df
def load_polarity_model(model_variant: str, base_dir: Path, device: str = "cuda"):
"""
Factory function to load polarity model by variant name.
Supported variants:
- "scibert": scibert/scibert_polarity/final_model (F1=0.724 baseline)
- "deberta": training/outputs/deberta_polarity/final_model (F1=0.764, +5.5% - RECOMMENDED)
- "deberta_v3_small": training/outputs/deberta_v3_small_polarity/final_model (F1=0.754)
- "modernbert": training/outputs/modernbert_polarity/final_model (F1=0.741)
- "scideberta": training/outputs/scideberta_polarity/final_model (F1=0.737)
Args:
model_variant: Name of model variant
base_dir: Base directory of project
device: Device to load onto
Returns:
Tuple of (tokenizer, model, device_obj)
Raises:
ValueError: If model_variant not supported
FileNotFoundError: If model directory doesn't exist
"""
# Feb 2026: New trained models from training/outputs/ (standardized comparison)
variant_map = {
"scibert": base_dir / "training" / "outputs" / "scibert_polarity" / "final_model",
"deberta": base_dir / "training" / "outputs" / "deberta_polarity" / "final_model", # BEST: F1=0.764
"deberta_v3_small": base_dir / "training" / "outputs" / "deberta_v3_small_polarity" / "final_model",
"modernbert": base_dir / "training" / "outputs" / "modernbert_polarity" / "final_model",
"scideberta": base_dir / "training" / "outputs" / "scideberta_polarity" / "final_model",
# Legacy models (pre-Feb 2026, kept for backwards compatibility)
"scibert_legacy": base_dir / "scibert" / "scibert_polarity" / "final_model",
"deberta_legacy": base_dir / "alternative_polarity" / "deberta" / "deberta_v3_base_polarity_final_model",
"scideberta_legacy": base_dir / "alternative_polarity" / "scideberta" / "scideberta_full_polarity_final_model",
}
hub_fallback_map = {
"scibert": "Sina1138/Scibert_polarity_Review",
"scideberta": "KISTI-AI/Scideberta-full", # Needs fine-tuning
"modernbert": "answerdotai/ModernBERT-base", # Needs fine-tuning
"deberta": "Sina1138/deberta_polarity_Review", # DeBERTa-v3-base (F1=0.764)
"deberta_v3_small": "microsoft/deberta-v3-small", # Needs fine-tuning
}
if model_variant not in variant_map:
raise ValueError(
f"Unknown polarity model variant: {model_variant}. "
f"Supported: {list(variant_map.keys())}"
)
model_dir = variant_map[model_variant]
return load_model_and_tokenizer(model_dir, device, hub_fallback=hub_fallback_map.get(model_variant))
def load_topic_model(model_variant: str, base_dir: Path, device: str = "cuda"):
"""
Factory function to load topic model by variant name.
Supported variants:
- "scideberta": training/outputs/scideberta_topic/final_model (F1=0.478 - BEST, RECOMMENDED)
- "deberta": training/outputs/deberta_topic/final_model (F1=0.450)
- "scibert": training/outputs/scibert_topic/final_model (F1=0.442)
- "deberta_v3_small": training/outputs/deberta_v3_small_topic/final_model (F1=0.381)
- "modernbert": training/outputs/modernbert_topic/final_model (F1=0.376)
Args:
model_variant: Name of model variant
base_dir: Base directory of project
device: Device to load onto
Returns:
Tuple of (tokenizer, model, device_obj)
Raises:
ValueError: If model_variant not supported
FileNotFoundError: If model directory doesn't exist
"""
# Feb 2026: New trained models from training/outputs/ (standardized comparison)
variant_map = {
"scideberta": base_dir / "training" / "outputs" / "scideberta_topic" / "final_model", # BEST: F1=0.478
"deberta": base_dir / "training" / "outputs" / "deberta_topic" / "final_model",
"scibert": base_dir / "training" / "outputs" / "scibert_topic" / "final_model",
"deberta_v3_small": base_dir / "training" / "outputs" / "deberta_v3_small_topic" / "final_model",
"modernbert": base_dir / "training" / "outputs" / "modernbert_topic" / "final_model",
# Legacy models (pre-Feb 2026, kept for backwards compatibility)
"scibert_legacy": base_dir / "scibert" / "scibert_topic" / "final_model",
"deberta_legacy": base_dir / "alternative_topic" / "deberta" / "final_model",
"scideberta_legacy": base_dir / "alternative_topic" / "scideberta" / "final_model",
}
hub_fallback_map = {
"scideberta": "Sina1138/scideberta_topic_Review", # SciDeBERTa (F1=0.478)
"scibert": "allenai/scibert_scivocab_uncased", # Needs fine-tuning
"deberta": "microsoft/deberta-v3-base", # Needs fine-tuning
"deberta_v3_small": "microsoft/deberta-v3-small", # Needs fine-tuning
"modernbert": "answerdotai/ModernBERT-base", # Needs fine-tuning
}
if model_variant not in variant_map:
raise ValueError(
f"Unknown topic model variant: {model_variant}. "
f"Supported: {list(variant_map.keys())}"
)
model_dir = variant_map[model_variant]
return load_model_and_tokenizer(model_dir, device, hub_fallback=hub_fallback_map.get(model_variant))
# Topic label mapping
TOPIC_ID_TO_LABEL = {
0: "Substance",
1: "Clarity",
2: "Soundness/Correctness",
3: "Originality",
4: "Motivation/Impact",
5: "Meaningful Comparison",
6: "Replicability",
7: "NONE",
}
TOPIC_LABEL_TO_ID = {v: k for k, v in TOPIC_ID_TO_LABEL.items()}
|