Hi, when I started the training progress, I met an error about cuda. Could you please tell me which version of cuda do you use?

[elegentbamboo]

I use a single GPU RTX3080, and cuda10.1, cudnn7.6.3, and torch1.4.0 as you specify in requirements.txt.
When I ran train.py to train the transfuser model, an error occurred: RuntimeError: cuDNN error: CUDNN_STATUS_EXECUTION_FAILED.
After searching this error, I set torch.backends.cudnn.enabled=false, but another error came: RuntimeError: CUDA error: CUBLAS_STATUS_EXECUTION_FAILED when calling cublasSgemm( handle, opa, opb, m, n, k, &alpha, a, lda, b, ldb, &beta, c, ldc). I guess it is the problem of version.

[ap229997]

I'd suggest updating cuda, cudnn, and torch to the latest version. I've seen cases where older cuda versions give errors when running on RTX 3080.