cuDNN error: CUDNN_STATUS_EXECUTION_FAILED
Open
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 9k
- Forks
- 1.5k
- Avg merge
- 2d 4h
- Merged PRs (30d)
- 3
Description
I get the following error every time I try to do a forward call with apex:
---------------------------------------------------------------------------
RuntimeError Traceback (most recent call last)
<ipython-input-20-c83117740453> in <module>
1 #%%pixie_debugger
2 while True:
----> 3 train(verbose=False, optimize_memory=True, optimize_feature=False)
4 with open('temp/memory.pkl', 'wb') as f:
5 pickle.dump(net.memory_model.memory, f)
<ipython-input-19-7e6a3b51254d> in train(verbose, optimize_memory, optimize_feature)
11 optimizer_both.zero_grad()
12
---> 13 similarities = net(batch_data)
14
15 values, indices = similarities.max(1)
~/miniconda3/lib/python3.6/site-packages/torch/nn/modules/module.py in __call__(self, *input, **kwargs)
487 result = self._slow_forward(*input, **kwargs)
488 else:
--> 489 result = self.forward(*input, **kwargs)
490 for hook in self._forward_hooks.values():
491 hook_result = hook(self, input, result)
<ipython-input-13-fa199304f042> in forward(self, images)
23 queries = self.feature_model(images)
24 #print(queries)
---> 25 similarities = self.memory_model(queries)
26 # print(sorted(similarities, reverse=True))
27 return similarities
~/miniconda3/lib/python3.6/site-packages/torch/nn/modules/module.py in __call__(self, *input, **kwargs)
487 result = self._slow_forward(*input, **kwargs)
488 else:
--> 489 result = self.forward(*input, **kwargs)
490 for hook in self._forward_hooks.values():
491 hook_result = hook(self, input, result)
<ipython-input-12-ea8dad5c6180> in forward(self, queries)
44
45 def forward(self, queries):
---> 46 sim_vector = self.get_similarity_vectors(queries)
47 return sim_vector
<ipython-input-12-ea8dad5c6180> in get_similarity_vectors(self, queries)
39
40 def get_similarity_vectors(self, queries):
---> 41 similarity = self.apply_combined(queries, self.memory, self.head_model)
42 # print(similarity)
43 return nn.functional.log_softmax(similarity * 10000) # multiply because of rounding errors
<ipython-input-12-ea8dad5c6180> in apply_combined(self, x, y, func)
34 assert x.shape == y.shape
35
---> 36 res = func(x, y)
37 res = res.view(n, m)
38 return res
~/miniconda3/lib/python3.6/site-packages/torch/nn/modules/module.py in __call__(self, *input, **kwargs)
487 result = self._slow_forward(*input, **kwargs)
488 else:
--> 489 result = self.forward(*input, **kwargs)
490 for hook in self._forward_hooks.values():
491 hook_result = hook(self, input, result)
~/Projects/Personal/Kaggle/humpwin/pancho111203/siamese/model.py in forward(self, x, y)
131 out = nn.functional.relu(out, inplace=True)
132 out = out.permute((0, 3, 1, 2))
--> 133 out = self.conv2(out)
134 out = out.view(batch_size, n_features)
135
~/miniconda3/lib/python3.6/site-packages/torch/nn/modules/module.py in __call__(self, *input, **kwargs)
487 result = self._slow_forward(*input, **kwargs)
488 else:
--> 489 result = self.forward(*input, **kwargs)
490 for hook in self._forward_hooks.values():
491 hook_result = hook(self, input, result)
~/miniconda3/lib/python3.6/site-packages/torch/nn/modules/conv.py in forward(self, input)
318 def forward(self, input):
319 return F.conv2d(input, self.weight, self.bias, self.stride,
--> 320 self.padding, self.dilation, self.groups)
321
322
~/miniconda3/lib/python3.6/site-packages/apex-0.1-py3.6-linux-x86_64.egg/apex/amp/wrap.py in wrapper(*args, **kwargs)
24 args,
25 kwargs)
---> 26 return orig_fn(*new_args, **kwargs)
27 return wrapper
28
RuntimeError: cuDNN error: CUDNN_STATUS_EXECUTION_FAILED
CUDNN logs:
https://gist.github.com/pancho111203/3e91f0b46ab0be3b04f1edc9c1405684
Contributor guide
No contributing guide indexed for this repository
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start with the failing convolution path in the reported model.py stack trace and the wrapper in apex/amp/wrap.py, then review the linked cuDNN logs. Reproduce the forward call with the supplied training configuration and narrow the report to a minimal, actionable failure with a confirmed resolution.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python, pytorch
- Domain
- machine-learning
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100