File size: 15,216 Bytes
9144bfc |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 |
import os
import sys
import json
import argparse
from typing import List, Dict, Tuple, Union, Any
from pathlib import Path
import pandas as pd
from tqdm.auto import tqdm
import torch
from transformers import (
MBart50Tokenizer, #type: ignore
MBartForConditionalGeneration, #type: ignore
MT5ForConditionalGeneration, #type: ignore
MT5TokenizerFast, #type: ignore
)
from peft import PeftModel, PeftConfig
import evaluate
# Add parent directory to sys.path
sys.path.append(str(Path(__file__).resolve().parent.parent))
from models.rule_based_mt import TransferBasedMT
from models.statistical_mt import SMTExtended, LanguageModel
# Device configuration
DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
# Load configuration
with open("config.json", "r") as json_file:
CONFIG = json.load(json_file)
def parse_arguments() -> argparse.Namespace:
"""Parse command-line arguments."""
parser = argparse.ArgumentParser(description="Evaluate English-Vietnamese Machine Translation Models")
parser.add_argument("--test_file", type=str, default='data/test_cleaned_dataset.csv', help="Path to test CSV file")
parser.add_argument("--output_dir", type=str, default="results", help="Directory to save results")
return parser.parse_args()
class ModelLoader:
"""Handles loading of translation models."""
@staticmethod
def load_smt() -> None:
"""Load Statistical Machine Translation model."""
try:
smt = SMTExtended()
model_dir = "checkpoints"
if os.path.exists(model_dir) and os.path.isfile(os.path.join(model_dir, "phrase_table.pkl")):
print("Loading existing model...")
smt.load_model()
else:
print("Training new smt...")
stats = smt.train()
print(f"Training complete: {stats}")
print("SMT model loaded successfully!")
return smt
except Exception as e:
raise RuntimeError(f"Failed to load SMT model: {str(e)}")
@staticmethod
def load_mbart50() -> Tuple[MBartForConditionalGeneration, MBart50Tokenizer]:
"""Load fine-tuned MBart50 model and tokenizer."""
try:
model_config = CONFIG["mbart50"]["paths"]
model = MBartForConditionalGeneration.from_pretrained(model_config["base_model_name"])
peft_config = PeftConfig.from_pretrained(model_config["checkpoint_path"])
model = PeftModel.from_pretrained(model, model_config["checkpoint_path"])
tokenizer = MBart50Tokenizer.from_pretrained(model_config["checkpoint_path"])
model.eval()
print("Fine-tuned MBart50 loaded successfully!")
return model.to(DEVICE), tokenizer #type: ignore
except Exception as e:
raise RuntimeError(f"Failed to load fine-tuned MBart50 model: {str(e)}")
@staticmethod
def load_original_mbart50() -> Tuple[MBartForConditionalGeneration, MBart50Tokenizer]:
"""Load original MBart50 model and tokenizer."""
try:
model_name = "facebook/mbart-large-50-many-to-many-mmt"
model = MBartForConditionalGeneration.from_pretrained(model_name)
tokenizer = MBart50Tokenizer.from_pretrained(model_name)
model.eval()
print("Original MBart50 loaded successfully!")
return model.to(DEVICE), tokenizer #type: ignore
except Exception as e:
raise RuntimeError(f"Failed to load original MBart50 model: {str(e)}")
@staticmethod
def load_mt5() -> Tuple[MT5ForConditionalGeneration, MT5TokenizerFast]:
"""Load fine-tuned MT5 model and tokenizer."""
try:
model_config = CONFIG["mt5"]["paths"]
model = MT5ForConditionalGeneration.from_pretrained(model_config["base_model_name"])
peft_config = PeftConfig.from_pretrained(model_config["checkpoint_path"])
model = PeftModel.from_pretrained(model, model_config["checkpoint_path"])
tokenizer = MT5TokenizerFast.from_pretrained(model_config["checkpoint_path"])
model.eval()
print("Fine-tuned MT5 loaded successfully!")
return model.to(DEVICE), tokenizer #type: ignore
except Exception as e:
raise RuntimeError(f"Failed to load fine-tuned MT5 model: {str(e)}")
@staticmethod
def load_original_mt5() -> Tuple[MT5ForConditionalGeneration, MT5TokenizerFast]:
"""Load original MT5 model and tokenizer."""
try:
model_name = "google/mt5-base"
model = MT5ForConditionalGeneration.from_pretrained(model_name)
tokenizer = MT5TokenizerFast.from_pretrained(model_name)
model.eval()
print("Original MT5 loaded successfully!")
return model.to(DEVICE), tokenizer #type: ignore
except Exception as e:
raise RuntimeError(f"Failed to load original MT5 model: {str(e)}")
class Translator:
"""Handles translation using different models."""
@staticmethod
def translate_rbmt(text: str) -> str:
"""Translate using Rule-Based Machine Translation."""
try:
translator = TransferBasedMT()
return translator.translate(text)
except Exception as e:
raise RuntimeError(f"RBMT translation failed: {str(e)}")
@staticmethod
def translate_smt(text: str, smt) -> str:
"""Translate using Statistical Machine Translation."""
try:
# return smt.translate_sentence(text)
translation = smt.infer(text)
return translation
except Exception as e:
raise RuntimeError(f"SMT translation failed: {str(e)}")
@staticmethod
def translate_mbart50(
model: MBartForConditionalGeneration, tokenizer: MBart50Tokenizer, text: str
) -> str:
"""Translate using MBart50 model (fine-tuned or original)."""
try:
model_config = CONFIG["mbart50"]["args"]
tokenizer.src_lang = model_config["src_lang"]
inputs = tokenizer(text, return_tensors="pt", padding=True)
inputs = {key: value.to(DEVICE) for key, value in inputs.items()}
forced_bos_token_id = tokenizer.lang_code_to_id[model_config["tgt_lang"]]
translated_tokens = model.generate(
input_ids=inputs["input_ids"],
attention_mask=inputs["attention_mask"],
forced_bos_token_id=forced_bos_token_id,
max_length=128,
num_beams=5,
)
return tokenizer.decode(translated_tokens[0], skip_special_tokens=True)
except Exception as e:
raise RuntimeError(f"MBart50 translation failed: {str(e)}")
@staticmethod
def translate_mt5(
model: MT5ForConditionalGeneration, tokenizer: MT5TokenizerFast, text: str
) -> str:
"""Translate using MT5 model (fine-tuned or original)."""
try:
prefix = CONFIG["mt5"]["args"]["prefix"]
text = prefix + text
inputs = tokenizer(text, return_tensors="pt", padding=True)
inputs = {key: value.to(DEVICE) for key, value in inputs.items()}
translated_tokens = model.generate(
input_ids=inputs["input_ids"],
attention_mask=inputs["attention_mask"],
max_length=128,
num_beams=5,
)
return tokenizer.decode(translated_tokens[0], skip_special_tokens=True)
except Exception as e:
raise RuntimeError(f"MT5 translation failed: {str(e)}")
class Evaluator:
"""Handles evaluation of translation models."""
@staticmethod
def load_test_data(test_file: str) -> List[Dict[str, str]]:
"""Load test data from CSV file."""
try:
df = pd.read_csv(test_file)
df = df[:100000] # Limit to 100,000 rows
return [{"source": row["en"], "reference": row["vi"]} for _, row in df.iterrows()]
except Exception as e:
raise RuntimeError(f"Failed to load test data: {str(e)}")
@staticmethod
def compute_metrics(hypotheses: List[str], references: List[str], sources: List[str]) -> Dict[str, float]:
"""Compute translation evaluation metrics."""
try:
metrics = {}
bleu_metric = evaluate.load("sacrebleu")
meteor_metric = evaluate.load("meteor")
rouge_metric = evaluate.load("rouge")
comet_metric = evaluate.load("comet")
bertscore_metric = evaluate.load("bertscore")
# BLEU
metrics["SacreBLEU"] = bleu_metric.compute(predictions=hypotheses, references=references)["score"] / 100 #type: ignore
# METEOR
metrics["METEOR"] = meteor_metric.compute(predictions=hypotheses, references=references)["meteor"] #type: ignore
# ROUGE
rouge_results = rouge_metric.compute(
predictions=hypotheses, references=references, rouge_types=["rouge1", "rouge2", "rougeL"], use_stemmer=True
)
metrics["ROUGE-1"] = rouge_results["rouge1"] #type: ignore
metrics["ROUGE-2"] = rouge_results["rouge2"] #type: ignore
metrics["ROUGE-L"] = rouge_results["rougeL"] #type: ignore
# BERTScore
bertscore_results = bertscore_metric.compute(
predictions=hypotheses, references=references, model_type="bert-base-multilingual-cased", lang="vi"
)
metrics["BERTScore"] = sum(bertscore_results["f1"]) / len(bertscore_results["f1"]) #type: ignore
# COMET
comet_results = comet_metric.compute(predictions=hypotheses, references=references, sources=sources)
metrics["COMET"] = sum(comet_results["scores"]) / len(comet_results["scores"]) #type: ignore
return metrics
except Exception as e:
raise RuntimeError(f"Failed to compute metrics: {str(e)}")
@staticmethod
def evaluate_model(
model_type: str, test_data: List[Dict[str, str]]
) -> Tuple[List[str], List[str], Dict[str, float]]:
"""Evaluate a translation model on test data."""
hypotheses, references, sources = [], [], []
try:
if model_type == "rbmt":
for item in tqdm(test_data, desc="Translating with RBMT"):
translation = Translator.translate_rbmt(item["source"])
hypotheses.append(translation)
references.append(item["reference"])
sources.append(item["source"])
elif model_type == "smt":
for item in tqdm(test_data, desc="Translating with SMT"):
smt = ModelLoader.load_smt()
translation = Translator.translate_smt(item["source"], smt)
hypotheses.append(translation)
references.append(item["reference"])
sources.append(item["source"])
elif model_type == "mbart50":
model, tokenizer = ModelLoader.load_mbart50()
for item in tqdm(test_data, desc="Translating with fine-tuned mBART50"):
translation = Translator.translate_mbart50(model, tokenizer, item["source"])
hypotheses.append(translation)
references.append(item["reference"])
sources.append(item["source"])
elif model_type == "original_mbart50":
model, tokenizer = ModelLoader.load_original_mbart50()
for item in tqdm(test_data, desc="Translating with original mBART50"):
translation = Translator.translate_mbart50(model, tokenizer, item["source"])
hypotheses.append(translation)
references.append(item["reference"])
sources.append(item["source"])
elif model_type == "mt5":
model, tokenizer = ModelLoader.load_mt5()
for item in tqdm(test_data, desc="Translating with fine-tuned MT5"):
translation = Translator.translate_mt5(model, tokenizer, item["source"])
hypotheses.append(translation)
references.append(item["reference"])
sources.append(item["source"])
elif model_type == "original_mt5":
model, tokenizer = ModelLoader.load_original_mt5()
for item in tqdm(test_data, desc="Translating with original MT5"):
translation = Translator.translate_mt5(model, tokenizer, item["source"])
hypotheses.append(translation)
references.append(item["reference"])
sources.append(item["source"])
return hypotheses, references, Evaluator.compute_metrics(hypotheses, references, sources) if hypotheses else {}
except Exception as e:
raise RuntimeError(f"Evaluation failed for {model_type}: {str(e)}")
def main():
"""Main function to run model evaluation."""
args = parse_arguments()
Path(args.output_dir).mkdir(parents=True, exist_ok=True)
try:
test_data = Evaluator.load_test_data(args.test_file)
model_types = ["rbmt", "smt" "mbart50", "original_mbart50", "mt5", "original_mt5"]
all_results = {}
for model_type in model_types:
print(f"\nEvaluating {model_type}...")
hypotheses, references, metrics = Evaluator.evaluate_model(model_type, test_data)
if metrics:
all_results[model_type] = metrics
print(f"Metrics for {model_type}:")
for metric, value in metrics.items():
print(f"{metric}: {value:.4f}")
# Save translations
translations = [
{"source": item["source"], "reference": ref, "hypothesis": hyp}
for item, ref, hyp in zip(test_data, references, hypotheses)
]
with open(
Path(args.output_dir) / f"{model_type}_translations.json", "w", encoding="utf-8"
) as f:
json.dump(translations, f, ensure_ascii=False, indent=2)
# Save all metrics
with open(Path(args.output_dir) / "metrics.json", "w", encoding="utf-8") as f:
json.dump(all_results, f, indent=2)
except Exception as e:
print(f"Error: {str(e)}", file=sys.stderr)
sys.exit(1)
if __name__ == "__main__":
main() |