Why Val loss is not showing ? how to display it then plot it with training loss

25 views Asked by At

after fine tune the pytorch llm(IDEFICS9b) on the data set the train results is not showing val loss how to collect it and then plot it with training loss?

training_args = TrainingArguments(
    output_dir=f"{model_name}-vqa1",
    learning_rate=3e-5,
    fp16=True,
    per_device_train_batch_size=16,
    per_device_eval_batch_size=16,
    gradient_accumulation_steps=1,
    dataloader_pin_memory=False,
    save_total_limit=3,
    evaluation_strategy="steps",
    save_strategy="steps",
    save_steps=10,
    eval_steps=10,
    logging_steps=10,
    max_steps=1000,
    remove_unused_columns=False,
    push_to_hub=False,
    label_names=["labels"],
    load_best_model_at_end=True,
    report_to=None,
    optim="paged_adamw_8bit",
)
trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=train_ds,
    eval_dataset=eval_ds,
)

train_results = trainer.train()

the train results is as follows:

TrainOutput(global_step=10, training_loss=2.2117252349853516, metrics={'train_runtime': 15.7995, 'train_samples_per_second': 10.127, 'train_steps_per_second': 0.633, 'total_flos': 470670709819392.0, 'train_loss': 2.2117252349853516, 'epoch': 0.02})

0

There are 0 answers