"No log" for training loss

🤗Transformers
1

training CLM task on Qwen3-0.6B-Base, here is validation Loss but no training loss, as shown below.

image
image291Ă—129 1.82 KB

Below are my training args:

from transformers import  TrainingArguments, Trainer

training_args = TrainingArguments(
    output_dir="CLM-Qwen3-0.6B-Base",
    eval_strategy="epoch",
    save_strategy="epoch",
#     per_device_train_batch_size=32,
#     per_device_eval_batch_size=32,
    learning_rate=2e-5,
    weight_decay=0.01,
    num_train_epochs=10,
    load_best_model_at_end=True,
    push_to_hub=False,
    save_total_limit=1,
    fp16=True,
)

trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=tokenized_datasets["train"],
    eval_dataset=tokenized_datasets["test"],
    processing_class=tokenizer,
    data_collator=data_collator,
)

trainer.train()```
1 Like