test_arch.py:
import torch
import torch.nn as nn
class net_lstm(nn.Module):
def __init__(self):
super(net_lstm, self).__init__()
self.lstm = nn.LSTM(10, 10)
def forward(self, input):
return input
class net_fc(nn.Module):
def __init__(self):
super(net_fc, self).__init__()
self.fc = nn.Linear(10, 10)
def forward(self, input):
return input
The following code for Linear layer works perfectly:
>>> import test_arch
>>> model_fc = test_arch.net_fc()
>>> model_fc
net_fc(
(fc): Linear(in_features=10, out_features=10, bias=True)
)
>>> model_fc.cuda()
net_fc(
(fc): Linear(in_features=10, out_features=10, bias=True)
)
But following for LSTM layer raises error:
>>> model_lstm = test_arch.net_lstm()
>>> model_lstm
net_lstm(
(lstm): LSTM(10, 10)
)
>>> model_lstm.cuda()
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/home/cse/ug/14074017/anaconda3/lib/python3.6/site-packages/torch/nn/modules/module.py", line 216, in cuda
return self._apply(lambda t: t.cuda(device))
File "/home/cse/ug/14074017/anaconda3/lib/python3.6/site-packages/torch/nn/modules/module.py", line 146, in _apply
module._apply(fn)
File "/home/cse/ug/14074017/anaconda3/lib/python3.6/site-packages/torch/nn/modules/rnn.py", line 123, in _apply
self.flatten_parameters()
File "/home/cse/ug/14074017/anaconda3/lib/python3.6/site-packages/torch/nn/modules/rnn.py", line 85, in flatten_parameters
handle = cudnn.get_handle()
File "/home/cse/ug/14074017/anaconda3/lib/python3.6/site-packages/torch/backends/cudnn/__init__.py", line 296, in get_handle
handle = CuDNNHandle()
File "/home/cse/ug/14074017/anaconda3/lib/python3.6/site-packages/torch/backends/cudnn/__init__.py", line 110, in __init__
check_error(lib.cudnnCreate(ctypes.byref(ptr)))
File "/home/cse/ug/14074017/anaconda3/lib/python3.6/site-packages/torch/backends/cudnn/__init__.py", line 283, in check_error
raise CuDNNError(status)
torch.backends.cudnn.CuDNNError: 4: b'CUDNN_STATUS_INTERNAL_ERROR'
>>>
Exception ignored in: <bound method CuDNNHandle.__del__ of <torch.backends.cudnn.CuDNNHandle object at 0x7f0a2230fba8>>
Traceback (most recent call last):
File "/home/cse/ug/14074017/anaconda3/lib/python3.6/site-packages/torch/backends/cudnn/__init__.py", line 114, in __del__
check_error(lib.cudnnDestroy(self))
ctypes.ArgumentError: argument 1: <class 'TypeError'>: Don't know how to convert parameter 1
KeyboardInterrupt
I tried updating to latest PyTorch version 0.3.1.post2, deleting ~/.nv and CUDA_CACHE_PATH='/home/your-unixname/cudacache' python main.py.
Please help.