Excuse me, can I use transformer-DLCL like transformer , how to operate it concretely, and whether I can call it directly? My code is tensorflow

@wangqiangneu 学长您好，我请教您一个问题。我在使用train-wmt-en2de-deep-prenorm-dlcl.sh进行训练时，会出现下面这个问题。按照错误提示我在_forward()里没有找到在哪里改正，也没见requires_grad在哪里，所以请问我需要怎么设置才能正常训练呢？

Traceback (most recent call last):
File "/home/zwc/python-virtual-environments/dlcl-master/multiprocessing_train.py", line 46, in run
single_process_main(args)
File "/home/zwc/python-virtual-environments/dlcl-master/train.py", line 126, in main
train(args, trainer, task, epoch_itr)
File "/home/zwc/python-virtual-environments/dlcl-master/train.py", line 172, in train
log_output = trainer.train_step(sample, update_params=True)
File "/home/zwc/python-virtual-environments/dlcl-master/fairseq/trainer.py", line 152, in train_step
grad_norm = self._all_reduce_and_rescale(grad_denom)
File "/home/zwc/python-virtual-environments/dlcl-master/fairseq/trainer.py", line 228, in _all_reduce_and_rescale
flat_grads = self._flat_grads = self._get_flat_grads(self._flat_grads)
File "/home/zwc/python-virtual-environments/dlcl-master/fairseq/trainer.py", line 253, in _get_flat_grads
grads = self._get_grads()
File "/home/zwc/python-virtual-environments/dlcl-master/fairseq/trainer.py", line 247, in _get_grads
raise RuntimeError('Model parameter did not receive gradient: ' + name + '. '
RuntimeError: Model parameter did not receive gradient: encoder.embed_tokens.weight. Use the param in the forward pass or set requires_grad=False

hello, i wonder if you could share your paper, the url is down
thanks

Can you provide the code and pre-trained model for English-Chinese translation?

Hey man,

I don't understand the function of inspect_grad. It doesn't seem to be used in your running script.
Could you please to explain it?

Thanks!

你好，数据集的链接好像失效了，请问可以重新生成一个链接吗？谢谢！