Hi, there
I currently ran into a problem below:
THCudaCheck FAIL file=/opt/v0.4.0/aten/src/THC/THCGeneral.cpp line=702 error=48 : no kernel image is available for execution on the device
Traceback (most recent call last):
File “/var/dl/runtime/script/main.py”, line 102, in
train(model, train_loader, eval_loader, args.epochs, args.output, optim, epoch)
File “/var/dl/runtime/script/tesla_test.zip/train.py”, line 94, in train
File “/opt/conda/envs/pytorch-py3.6/lib/python3.6/site-packages/torch/nn/modules/module.py”, line 491, in call
result = self.forward(*input, **kwargs)
File “/var/dl/runtime/script/tesla_test.zip/test_base_model.py”, line 48, in forward
File “/opt/conda/envs/pytorch-py3.6/lib/python3.6/site-packages/torch/nn/modules/module.py”, line 491, in call
result = self.forward(*input, **kwargs)
File “/var/dl/runtime/script/tesla_test.zip/language_model.py”, line 43, in forward
RuntimeError: cuda runtime error (48) : no kernel image is available for execution on the device at /opt/v0.4.0/aten/src/THC/THCGeneral.cpp:702
When the exact same code ran on local, which is cuda 8.0 environment. It can finish without any error.
However, when I ran on platform, which is cuda 9.0. It gave me such error.
I was wondering what is the difference between the program that can run on 8.0 or 9.0. Don’t it use the same api?
I’am here to ask how can I find the changes that needed to be apply to my program so that it is compatible to 9.0