Avoid printing 'Generate config GenerationConfig { ... }'

Question

I am facing an issue while training a t5 model. After each evaluation step, the following message is printed, which makes it impossible to maintain an overview. Do you have any ideas, on how I can avoid such behavior?

***** Running Evaluation ***** Num examples = 819 Batch size = 32 Generate config GenerationConfig { "decoder_start_token_id": 0,
"eos_token_id": 1, "output_attentions": true,
"output_hidden_states": true, "pad_token_id": 0,
"transformers_version": "4.26.1" }

Generate config GenerationConfig { "decoder_start_token_id": 0,
"eos_token_id": 1, "output_attentions": true,
"output_hidden_states": true, "pad_token_id": 0,
"transformers_version": "4.26.1" }

Generate config GenerationConfig { "decoder_start_token_id": 0,
"eos_token_id": 1, "output_attentions": true,
"output_hidden_states": true, "pad_token_id": 0,
"transformers_version": "4.26.1" } ...

from transformers import AutoModelForSeq2SeqLM

model_id="google/flan-t5-base"
model = AutoModelForSeq2SeqLM.from_pretrained(model_id)

repository_id = f"{model_id.split('/')[1]}-{dataset_id}"

training_args = Seq2SeqTrainingArguments(
    output_dir=repository_id,
    #gradient_accumulation_steps = 8,
    per_device_train_batch_size=8,
    per_device_eval_batch_size=8,
    predict_with_generate=True,
    fp16=False, # Overflows with fp16
    learning_rate=5e-6,
    num_train_epochs=5,
    optim = "adamw_torch",
    logging_dir=f"{repository_id}/logs",
    logging_strategy="steps",
    logging_steps=50,
    evaluation_strategy="steps",
    eval_steps=5,
    save_strategy="steps",
    save_total_limit=2,
    load_best_model_at_end=True,
    report_to="tensorboard",
    push_to_hub=False,
    hub_strategy="every_save",
    hub_model_id=repository_id,
    hub_token=HfFolder.get_token(),
)

trainer = Seq2SeqTrainer(
    model=model,
    args=training_args,
    data_collator=data_collator,
    train_dataset=tokenized_dataset["train"],
    eval_dataset=tokenized_dataset["test"],
    compute_metrics=compute_metrics,
)

trainer.train()

I tried to change the log level, but that does not help.

import os
os.environ.TF_CPP_MIN_LOG_LEVEL=2

Avoid printing 'Generate config GenerationConfig { ... }'

0 Answers0