Closed dongzhuoyao closed 2 months ago
Hi, maybe you should try sth like:
o.backward(torch.randn_like(o))
This error seems like an OOM error.
thanks for your reply, it doens't solve this issue, I somehow figure out the issue is here:
but has no clue to how to solve it
Will this error still happen if you reduce the batch size or img size?
yes, it still have, and the gpu is only occupied by me(A100)
Thanks for your comment. I'll tell you if I figure it out later.
python3.11, cuda11.8, torch2.2.0
For RWKV4, I can run it successfully
Hi, maybe increase T_MAX
in vrwkv6.py can solve this.
thanks, increase T_max works for me
Hi, I did the following simple test for RWKV6, but it shows the following error.
Have you met this before? could you share me your detailed python,cuda,torch version?
Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass RuntimeError: CUDA error: an illegal memory access was encountered CUDA kernel errors might be asynchronously reported at some other API call,so the stacktrace below might be incorrect. For debugging consider passing CUDA_LAUNCH_BLOCKING=1.