lhyscau

the json file (attempt/configs/attempt/eval_glue.json) may forget to add the dataset stsb and cola in the paper？And the dataset wnli which is not in the paper and I think that the...

关于代码model.enable_input_require_grads()

4

` if training_args.do_train: model.gradient_checkpointing_enable() model.enable_input_require_grads()` 作者你好，请问这里令输入也计算梯度的目的是什么呢？是否冗余了这条代码，还是说这个一个涨分的方式？

In the multi-task target traing, the file "multi_task_superglue.json" should set the param "shared_attn" true, otherwise the train dataset didn't have the task_ids.