Bert Trainer 实例中的提前停止

Question

我正在为多类分类任务微调 BERT 模型。我的问题是我不知道如何为那些 Trainer 实例添加“提前停止”。有什么想法吗？

Answer 1

在正确使用 EarlyStoppingCallback()

之前，您需要进行一些修改

from transformers import EarlyStoppingCallback
...
...
# Defining the TrainingArguments() arguments
args = TrainingArguments(
   f"training_with_callbacks",
   evaluation_strategy ='steps',
   eval_steps = 50, # Evaluation and Save happens every 50 steps
   save_total_limit = 5, # Only last 5 models are saved. Older ones are deleted.
   learning_rate=2e-5,
   per_device_train_batch_size=batch_size,
   per_device_eval_batch_size=batch_size,
   num_train_epochs=5,
   weight_decay=0.01,
   push_to_hub=False,
   metric_for_best_model = 'f1',
   load_best_model_at_end=True)

您需要：

使用load_best_model_at_end = True（EarlyStoppingCallback()要求为True）。
evaluation_strategy = 'steps' 而不是 'epoch'.
eval_steps = 50（评估 N 步后的指标）。
metric_for_best_model = 'f1',

在你的 Trainer():

trainer = Trainer(
    model,
    args,
    ...
    compute_metrics=compute_metrics,
    callbacks = [EarlyStoppingCallback(early_stopping_patience=3)]
)

当然，当你使用compute_metrics时，例如它可以是这样的函数：

def compute_metrics(p):    
    pred, labels = p
    pred = np.argmax(pred, axis=1)
    accuracy = accuracy_score(y_true=labels, y_pred=pred)
    recall = recall_score(y_true=labels, y_pred=pred)
    precision = precision_score(y_true=labels, y_pred=pred)
    f1 = f1_score(y_true=labels, y_pred=pred)    
return {"accuracy": accuracy, "precision": precision, "recall": recall, "f1": f1}

compute_metrics() 的 return 应该是字典，您可以在函数内访问 want/compute 和 return.

的任何指标

Bert Trainer 实例中的提前停止

Early stopping in Bert Trainer instances

python

neural-network

huggingface-transformers

bert-language-model