search

INK-USC / RE-Net

Recurrent Event Network: Autoregressive Structure Inference over Temporal Knowledge Graphs (EMNLP 2020)
http://inklab.usc.edu/renet/
436 stars 95 forks source link

CUDA error: no kernel image is available for execution on the device #77

Open Sitiange opened 8 months ago

Sitiange commented 8 months ago

I have configured the environment as required, but I still get a CUDA error when running pretrain.py, why?

Environment: python 3.6 torch 1.6.0+cu101 dgl-cu101 0.4.3 CUDA 10.1 nvcc: NVIDIA (R) Cuda compiler driver Copyright (c) 2005-2019 NVIDIA Corporation Built on Sun_Jul_28_19:07:16_PDT_2019 Cuda compilation tools, release 10.1, V10.1.243

Traceback (most recent call last): File "pretrain.py", line 139, in train(args) File "pretrain.py", line 51, in train model.cuda() File "/home/sitiange/miniconda3/envs/renet/lib/python3.6/site-packages/torch/nn/modules/module.py", line 458, in cuda return self._apply(lambda t: t.cuda(device)) File "/home/sitiange/miniconda3/envs/renet/lib/python3.6/site-packages/torch/nn/modules/module.py", line 354, in _apply module._apply(fn) File "/home/sitiange/miniconda3/envs/renet/lib/python3.6/site-packages/torch/nn/modules/rnn.py", line 161, in _apply self.flatten_parameters() File "/home/sitiange/miniconda3/envs/renet/lib/python3.6/site-packages/torch/nn/modules/rnn.py", line 151, in flatten_parameters self.batch_first, bool(self.bidirectional)) RuntimeError: CUDA error: no kernel image is available for execution on the device

Sitiange commented 8 months ago

I find the reason, NVIDIA A40 with CUDA capability sm_86 is not compatible with PyTorch 1.6...