Haha, I use the GPU on my laptop, but I have to set BATCH_SIZE to 1 so I don't run out of memory，there are too many Total params about casrel

[Ding-Chian]

但是你可以在model.py中更改以下代码
for l in bert_model.layers:
l.trainable = True

#只是微调后8层（举个例子）
trainable_layers = 8 # 后八层
num_layers = len(bert_model.layers)
for i, layer in enumerate(bert_model.layers):
if i >= num_layers - trainable_layers:
layer.trainable = True
else:
layer.trainable = False
从而减少参数数量