Transformer模型报错

[bitsk]

W0915 12:02:58.731670 1250 gpu_resources.cc:61] Please NOTE: device: 0, GPU Compute Capability: 8.6, Driver API Version: 11.7, Runtime API Version: 11.7 W0915 12:02:58.744315 1250 gpu_resources.cc:91] device: 0, cuDNN Version: 8.5. Traceback (most recent call last): File "test.py", line 34, in model.fit(train_dataset, val_dataset) File "/usr/local/lib/python3.8/dist-packages/paddlets/models/dl/paddlepaddle/paddle_base_impl.py", line 321, in fit self._fit(train_dataloader, valid_dataloaders) File "/usr/local/lib/python3.8/dist-packages/paddlets/models/dl/paddlepaddle/paddle_base_impl.py", line 347, in _fit self._train_epoch(train_dataloader) File "/usr/local/lib/python3.8/dist-packages/paddlets/models/dl/paddlepaddle/paddle_base_impl.py", line 415, in _train_epoch batch_logs = self._train_batch(X, y) File "/usr/local/lib/python3.8/dist-packages/paddlets/models/dl/paddlepaddle/paddle_base_impl.py", line 434, in _train_batch output = self._network(X) File "/usr/local/lib/python3.8/dist-packages/paddle/fluid/dygraph/layers.py", line 930, in call return self._dygraph_call_func(*inputs, **kwargs) File "/usr/local/lib/python3.8/dist-packages/paddle/fluid/dygraph/layers.py", line 915, in _dygraph_call_func outputs = self.forward(*inputs, **kwargs) File "/usr/local/lib/python3.8/dist-packages/paddlets/models/dl/paddlepaddle/transformer.py", line 196, in forward out = self._transformer(src, tgt) File "/usr/local/lib/python3.8/dist-packages/paddle/fluid/dygraph/layers.py", line 930, in call return self._dygraph_call_func(*inputs, **kwargs) File "/usr/local/lib/python3.8/dist-packages/paddle/fluid/dygraph/layers.py", line 915, in _dygraph_call_func outputs = self.forward(*inputs, **kwargs) File "/usr/local/lib/python3.8/dist-packages/paddle/nn/layer/transformer.py", line 1628, in forward memory = self.encoder(src, src_mask=src_mask) File "/usr/local/lib/python3.8/dist-packages/paddle/fluid/dygraph/layers.py", line 930, in call return self._dygraph_call_func(*inputs, **kwargs) File "/usr/local/lib/python3.8/dist-packages/paddle/fluid/dygraph/layers.py", line 915, in _dygraph_call_func outputs = self.forward(*inputs, **kwargs) File "/usr/local/lib/python3.8/dist-packages/paddle/nn/layer/transformer.py", line 971, in forward src_mask = _convert_attention_mask(src_mask, src.dtype, File "/usr/local/lib/python3.8/dist-packages/paddle/nn/layer/transformer.py", line 108, in _convert_attention_mask mha_meta = _prepare_mha_meta(attn_mask, enable_cudnn) File "/usr/local/lib/python3.8/dist-packages/paddle/nn/layer/transformer.py", line 134, in _prepare_mha_meta assert attn_mask is not None,
AssertionError: The attention mask should be given for MultiHeadAttention when enable_cudnn=True. But received attn_mask = None

[willionZS]

看起来是attn_mask这个参数为None的情况造成的？

[Annnnnnnnnnnnn]

能提供一下paddle的版本么，最好可以提供一份可以复现的代码最小集合。