报错
Input, output and indices must be on the current device
检查模型需要的输入是否放在gpu上
ids = batch['input_ids'].to(device)
types = batch['token_type_ids'].to(device)
mask = batch['attention_mask'].to(device)
labels = batch['labels'].to(device)
还有非常重要的一点
👉 模型也要放到GPU上
model.to(device)