| 0%| | 0/892 [00:00<?, ?it/s][WARNING|modeling_utils.py:388] 2022-03-03 00:40:45,319 >> Could not estimate the number of tokens of the input, floating-point operations will not be computed | |
| Traceback (most recent call last): | |
| File "/home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/run_speech_recognition_seq2seq.py", line 539, in <module> | |
| main() | |
| File "/home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/run_speech_recognition_seq2seq.py", line 491, in main | |
| train_result = trainer.train(resume_from_checkpoint=checkpoint) | |
| File "/home/sanchit_huggingface_co/transformers/src/transformers/trainer.py", line 1384, in train | |
| tr_loss_step = self.training_step(model, inputs) | |
| File "/home/sanchit_huggingface_co/transformers/src/transformers/trainer.py", line 1959, in training_step | |
| loss = self.compute_loss(model, inputs) | |
| File "/home/sanchit_huggingface_co/transformers/src/transformers/trainer.py", line 1991, in compute_loss | |
| outputs = model(**inputs) | |
| File "/home/sanchit_huggingface_co/gcp/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl | |
| return forward_call(*input, **kwargs) | |
| File "/home/sanchit_huggingface_co/transformers/src/transformers/models/speech_encoder_decoder/modeling_speech_encoder_decoder.py", line 503, in forward | |
| encoder_outputs = self.encoder( | |
| File "/home/sanchit_huggingface_co/gcp/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl | |
| return forward_call(*input, **kwargs) | |
| File "/home/sanchit_huggingface_co/transformers/src/transformers/models/wav2vec2/modeling_wav2vec2.py", line 1346, in forward | |
| extract_features = self.feature_extractor(input_values) | |
| File "/home/sanchit_huggingface_co/gcp/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl | |
| return forward_call(*input, **kwargs) | |
| File "/home/sanchit_huggingface_co/transformers/src/transformers/models/wav2vec2/modeling_wav2vec2.py", line 514, in forward | |
| hidden_states = conv_layer(hidden_states) | |
| File "/home/sanchit_huggingface_co/gcp/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl | |
| return forward_call(*input, **kwargs) | |
| File "/home/sanchit_huggingface_co/transformers/src/transformers/models/wav2vec2/modeling_wav2vec2.py", line 389, in forward | |
| hidden_states = self.layer_norm(hidden_states) | |
| File "/home/sanchit_huggingface_co/gcp/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl | |
| return forward_call(*input, **kwargs) | |
| File "/home/sanchit_huggingface_co/gcp/lib/python3.9/site-packages/torch/nn/modules/normalization.py", line 189, in forward | |
| return F.layer_norm( | |
| File "/home/sanchit_huggingface_co/gcp/lib/python3.9/site-packages/torch/nn/functional.py", line 2347, in layer_norm | |
| return torch.layer_norm(input, normalized_shape, weight, bias, eps, torch.backends.cudnn.enabled) | |
| RuntimeError: CUDA out of memory. Tried to allocate 1.63 GiB (GPU 0; 15.78 GiB total capacity; 10.07 GiB already allocated; 707.31 MiB free; 13.39 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF |