423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-12-16 11:50:35 +08:00
Code Issues Packages Projects Releases Wiki Activity

modify style

Browse Source
This commit is contained in:
BUAADreamer
2024-04-25 21:15:16 +08:00
parent 43d7ad5ecc
commit 1dcabafe72
16 changed files with 374 additions and 502 deletions
Download Patch File Download Diff File Expand all files Collapse all files

50
src/llmtuner/train/sft/workflow.py
View File

@@ -50,29 +50,17 @@ def run_sft(
tokenizer.padding_side = "left" # use left-padding in generation
if getattr(model, "is_quantized", False) and not training_args.do_train:
setattr(
model, "_hf_peft_config_loaded", True
) # hack here: make model compatible with prediction
setattr(model, "_hf_peft_config_loaded", True) # hack here: make model compatible with prediction
data_collator = DataCollatorForSeq2Seq(
tokenizer=tokenizer,
pad_to_multiple_of=(
8 if tokenizer.padding_side == "right" else None
), # for shift short attention
label_pad_token_id=(
IGNORE_INDEX
if data_args.ignore_pad_token_for_loss
else tokenizer.pad_token_id
),
pad_to_multiple_of=(8 if tokenizer.padding_side == "right" else None), # for shift short attention
label_pad_token_id=(IGNORE_INDEX if data_args.ignore_pad_token_for_loss else tokenizer.pad_token_id),
)
# Override the decoding parameters of Seq2SeqTrainer
training_args.generation_max_length = (
training_args.generation_max_length or data_args.cutoff_len
)
training_args.generation_num_beams = (
data_args.eval_num_beams or training_args.generation_num_beams
)
training_args.generation_max_length = training_args.generation_max_length or data_args.cutoff_len
training_args.generation_num_beams = data_args.eval_num_beams or training_args.generation_num_beams
if model_args.use_mllm:
training_args.remove_unused_columns = False
@@ -84,25 +72,19 @@ def run_sft(
tokenizer=tokenizer,
data_collator=data_collator,
callbacks=callbacks,
compute_metrics=(
ComputeMetrics(tokenizer) if training_args.predict_with_generate else None
),
compute_metrics=(ComputeMetrics(tokenizer) if training_args.predict_with_generate else None),
**split_dataset(dataset, data_args, training_args),
)
# Keyword arguments for `model.generate`
gen_kwargs = generating_args.to_dict()
gen_kwargs["eos_token_id"] = [
tokenizer.eos_token_id
] + tokenizer.additional_special_tokens_ids
gen_kwargs["eos_token_id"] = [tokenizer.eos_token_id] + tokenizer.additional_special_tokens_ids
gen_kwargs["pad_token_id"] = tokenizer.pad_token_id
gen_kwargs["logits_processor"] = get_logits_processor()
# Training
if training_args.do_train:
train_result = trainer.train(
resume_from_checkpoint=training_args.resume_from_checkpoint
)
train_result = trainer.train(resume_from_checkpoint=training_args.resume_from_checkpoint)
trainer.save_model()
trainer.log_metrics("train", train_result.metrics)
trainer.save_metrics("train", train_result.metrics)
@@ -113,27 +95,19 @@ def run_sft(
# Evaluation
if training_args.do_eval:
metrics = trainer.evaluate(metric_key_prefix="eval", **gen_kwargs)
if (
training_args.predict_with_generate
): # eval_loss will be wrong if predict_with_generate is enabled
if training_args.predict_with_generate: # eval_loss will be wrong if predict_with_generate is enabled
metrics.pop("eval_loss", None)
trainer.log_metrics("eval", metrics)
trainer.save_metrics("eval", metrics)
# Predict
if training_args.do_predict:
predict_results = trainer.predict(
dataset, metric_key_prefix="predict", **gen_kwargs
)
if (
training_args.predict_with_generate
): # predict_loss will be wrong if predict_with_generate is enabled
predict_results = trainer.predict(dataset, metric_key_prefix="predict", **gen_kwargs)
if training_args.predict_with_generate: # predict_loss will be wrong if predict_with_generate is enabled
predict_results.metrics.pop("predict_loss", None)
trainer.log_metrics("predict", predict_results.metrics)
trainer.save_metrics("predict", predict_results.metrics)
trainer.save_predictions(predict_results)
# Create model card
create_modelcard_and_push(
trainer, model_args, data_args, training_args, finetuning_args
)
create_modelcard_and_push(trainer, model_args, data_args, training_args, finetuning_args)