"""
Fine-tuning the library models for sequence to sequence.
"""
import logging
import os
import sys
import json
import numpy as np
from datasets import load_dataset, load_from_disk
import jieba
from rouge_chinese import Rouge
from nltk.translate.bleu_score import sentence_bleu, SmoothingFunction
import torch
import torch_npu
import deepspeed_npu
from torch_npu.contrib import transfer_to_npu
import transformers
from transformers import (
AutoConfig,
AutoModel,
AutoTokenizer,
DataCollatorForSeq2Seq,
HfArgumentParser,
Seq2SeqTrainingArguments,
set_seed,
)
from trainer_seq2seq import Seq2SeqTrainer
from arguments import ModelArguments, DataTrainingArguments
logger = logging.getLogger(__name__)
def main():
torch.use_deterministic_algorithms(True)
parser = HfArgumentParser((ModelArguments, DataTrainingArguments, Seq2SeqTrainingArguments))
if len(sys.argv) == 2 and sys.argv[1].endswith(".json"):
model_args, data_args, training_args = parser.parse_json_file(json_file=os.path.abspath(sys.argv[1]))
else:
model_args, data_args, training_args = parser.parse_args_into_dataclasses()
if training_args.do_train:
torch.npu.set_compile_mode(jit_compile=True)
if training_args.do_predict:
training_args.local_rank = -1
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
datefmt="%m/%d/%Y %H:%M:%S",
handlers=[logging.StreamHandler(sys.stdout)],
)
if training_args.should_log:
transformers.utils.logging.set_verbosity_info()
log_level = training_args.get_process_log_level()
logger.setLevel(log_level)
transformers.utils.logging.set_verbosity(log_level)
transformers.utils.logging.enable_default_handler()
transformers.utils.logging.enable_explicit_format()
logger.warning(
f"Process rank: {training_args.local_rank}, device: {training_args.device}, n_gpu: {training_args.n_gpu}"
+ f"distributed training: {bool(training_args.local_rank != -1)}, 16-bits training: {training_args.fp16}"
)
logger.info(f"Training/evaluation parameters {training_args}")
set_seed(training_args.seed)
config = AutoConfig.from_pretrained(model_args.model_name_or_path, trust_remote_code=True)
config.pre_seq_len = model_args.pre_seq_len
config.prefix_projection = model_args.prefix_projection
tokenizer = AutoTokenizer.from_pretrained(model_args.model_name_or_path, trust_remote_code=True)
if model_args.ptuning_checkpoint is not None:
model = AutoModel.from_pretrained(model_args.model_name_or_path, config=config, trust_remote_code=True)
prefix_state_dict = torch.load(os.path.join(model_args.ptuning_checkpoint, "pytorch_model.bin"))
new_prefix_state_dict = {}
for k, v in prefix_state_dict.items():
if k.startswith("transformer.prefix_encoder."):
new_prefix_state_dict[k[len("transformer.prefix_encoder."):]] = v
model.transformer.prefix_encoder.load_state_dict(new_prefix_state_dict)
else:
model = AutoModel.from_pretrained(model_args.model_name_or_path, config=config, trust_remote_code=True)
if model_args.quantization_bit is not None:
print(f"Quantized to {model_args.quantization_bit} bit")
model = model.quantize(model_args.quantization_bit)
if model_args.pre_seq_len is not None:
model = model.half()
model.transformer.prefix_encoder.float()
else:
model = model.float()
prefix = data_args.source_prefix if data_args.source_prefix is not None else ""
def print_dataset_example(example):
print("input_ids", example["input_ids"])
print("inputs", tokenizer.decode(example["input_ids"]))
print("label_ids", example["labels"])
print("labels", tokenizer.decode(example["labels"]))
if training_args.do_train:
train_dataset = load_from_disk("train_datasets")
print_dataset_example(train_dataset[0])
if data_args.max_train_samples is not None:
max_train_samples = min(len(train_dataset), data_args.max_train_samples)
train_dataset = train_dataset.select(range(max_train_samples))
if training_args.do_eval:
eval_dataset = load_from_disk("validation_datasets")
if data_args.max_eval_samples is not None:
max_eval_samples = min(len(eval_dataset), data_args.max_eval_samples)
eval_dataset = eval_dataset.select(range(max_eval_samples))
if training_args.do_predict:
predict_dataset = load_from_disk("test_datasets")
if data_args.max_predict_samples is not None:
max_predict_samples = min(len(predict_dataset), data_args.max_predict_samples)
predict_dataset = predict_dataset.select(range(max_predict_samples))
label_pad_token_id = -100 if data_args.ignore_pad_token_for_loss else tokenizer.pad_token_id
data_collator = DataCollatorForSeq2Seq(
tokenizer,
model=model,
label_pad_token_id=label_pad_token_id,
pad_to_multiple_of=None,
padding=False
)
def compute_metrics(eval_preds):
preds, labels = eval_preds
if isinstance(preds, tuple):
preds = preds[0]
decoded_preds = tokenizer.batch_decode(preds, skip_special_tokens=True)
if data_args.ignore_pad_token_for_loss:
labels = np.where(labels != -100, labels, tokenizer.pad_token_id)
decoded_labels = tokenizer.batch_decode(labels, skip_special_tokens=True)
score_dict = {
"rouge-1": [],
"rouge-2": [],
"rouge-l": [],
"bleu-4": []
}
for pred, label in zip(decoded_preds, decoded_labels):
hypothesis = list(jieba.cut(pred))
reference = list(jieba.cut(label))
rouge = Rouge()
scores = rouge.get_scores(' '.join(hypothesis), ' '.join(reference))
result = scores[0]
for k, v in result.items():
score_dict[k].append(round(v["f"] * 100, 4))
bleu_score = sentence_bleu([list(label)], list(pred), smoothing_function=SmoothingFunction().method3)
score_dict["bleu-4"].append(round(bleu_score * 100, 4))
for k, v in score_dict.items():
score_dict[k] = float(np.mean(v))
return score_dict
training_args.generation_max_length = (
training_args.generation_max_length
if training_args.generation_max_length is not None
else data_args.val_max_target_length
)
training_args.generation_num_beams = (
data_args.num_beams if data_args.num_beams is not None else training_args.generation_num_beams
)
trainer = Seq2SeqTrainer(
model=model,
args=training_args,
train_dataset=train_dataset if training_args.do_train else None,
eval_dataset=eval_dataset if training_args.do_eval else None,
tokenizer=tokenizer,
data_collator=data_collator,
compute_metrics=compute_metrics if training_args.predict_with_generate else None,
save_changed=model_args.pre_seq_len is not None
)
if training_args.do_train:
checkpoint = None
if training_args.resume_from_checkpoint is not None:
checkpoint = training_args.resume_from_checkpoint
model.gradient_checkpointing_enable()
model.enable_input_require_grads()
train_result = trainer.train(resume_from_checkpoint=checkpoint)
metrics = train_result.metrics
max_train_samples = (
data_args.max_train_samples if data_args.max_train_samples is not None else len(train_dataset)
)
metrics["train_samples"] = min(max_train_samples, len(train_dataset))
trainer.log_metrics("train", metrics)
trainer.save_metrics("train", metrics)
trainer.save_state()
results = {}
max_seq_length = data_args.max_source_length + data_args.max_target_length + 1
if training_args.do_eval:
logger.info("*** Evaluate ***")
metrics = trainer.evaluate(metric_key_prefix="eval", do_sample=True, top_p=0.7, max_length=max_seq_length,
temperature=0.95)
max_eval_samples = data_args.max_eval_samples if data_args.max_eval_samples is not None else len(eval_dataset)
metrics["eval_samples"] = min(max_eval_samples, len(eval_dataset))
trainer.log_metrics("eval", metrics)
trainer.save_metrics("eval", metrics)
if training_args.do_predict:
logger.info("*** Predict ***")
predict_results = trainer.predict(predict_dataset, metric_key_prefix="predict", max_length=max_seq_length,
do_sample=True, top_p=0.7, temperature=0.95)
metrics = predict_results.metrics
max_predict_samples = (
data_args.max_predict_samples if data_args.max_predict_samples is not None else len(predict_dataset)
)
metrics["predict_samples"] = min(max_predict_samples, len(predict_dataset))
trainer.log_metrics("predict", metrics)
trainer.save_metrics("predict", metrics)
if trainer.is_world_process_zero():
if training_args.predict_with_generate:
predictions = tokenizer.batch_decode(
predict_results.predictions, skip_special_tokens=True, clean_up_tokenization_spaces=True
)
predictions = [pred.strip() for pred in predictions]
labels = tokenizer.batch_decode(
predict_results.label_ids, skip_special_tokens=True, clean_up_tokenization_spaces=True
)
labels = [label.strip() for label in labels]
output_prediction_file = os.path.join(training_args.output_dir, "generated_predictions.txt")
with open(output_prediction_file, "w", encoding="utf-8") as writer:
for p, l in zip(predictions, labels):
res = json.dumps({"labels": l, "predict": p}, ensure_ascii=False)
writer.write(f"{res}\n")
return results
def _mp_fn(index):
main()
if __name__ == "__main__":
main()