search

Vincent131499 / TextSim_cn_finetune

微调预训练语言模型(BERT、Roberta、XLBert等),用于计算两个文本之间的相似度（通过句子对分类任务转换）,适用于中文文本
89 stars 20 forks source link

训练模型需要多久啊，这是正常情况嘛 #2

Open Oohuo opened 4 years ago

Oohuo commented 4 years ago

It seems that global step (tf.train.get_global_step) has not been increased. Current value (could be stable): 97 vs previous value: 97. You could increase the global step by passing tf.train.get_global_step() to Optimizer.apply_gradients or Optimizer.minimize.

0318 21:04:46.586845 140013223937088 tpu_estimator.py:2159] global_step/sec: 0.319628 INFO:tensorflow:examples/sec: 2.55703 I0318 21:04:46.587145 140013223937088 tpu_estimator.py:2160] examples/sec: 2.55703 INFO:tensorflow:global_step/sec: 0.30417 I0318 21:04:49.874457 140013223937088 tpu_estimator.py:2159] global_step/sec: 0.30417 INFO:tensorflow:examples/sec: 2.43336 I0318 21:04:49.874915 140013223937088 tpu_estimator.py:2160] examples/sec: 2.43336 INFO:tensorflow:global_step/sec: 0.285283 I0318 21:04:53.379786 140013223937088 tpu_estimator.py:2159] global_step/sec: 0.285283 INFO:tensorflow:examples/sec: 2.28227 I0318 21:04:53.380270 140013223937088 tpu_estimator.py:2160] examples/sec: 2.28227 INFO:tensorflow:global_step/sec: 0.291327

Vincent131499 commented 4 years ago

Hello，这种情况应该是训练速度过慢导致的，属于正常情况。你可以通过降低train_batch_size或者调整learning_rate来避免这个warning