torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 196.00 MiB. GPU 0 has a total capacty of 79.11 GiB of which 29.56 MiB is free

#21
by butujuzipi - opened

torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 196.00 MiB. GPU 0 has a total capacty of 79.11 GiB of which 29.56 MiB is free
How should I solve this problem?

DeepSeek org

You need to load the model with bf16

DeepSeek org

You can hardly run this model on a single GPU. Try a smaller model or use model parallelism to run it on multiple GPUs instead.

Chester111 changed discussion status to closed

Sign up or log in to comment