Skip to content

CUDA out of memory #4

Description

@maa989

I am using the default config, and also running on a system with 8 A100 GPUs (each 40GB vRAM). But i still get the following error:
torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 80.00 MiB. GPU 1 has a total capacity of 39.50 GiB of which 17.38 MiB is free. Including non-PyTorch memory, this process has 39.47 GiB memory in use. Of the allocated memory 39.06 GiB is allocated by PyTorch, and 8.91 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables)

How much memory is needed to run this? and is there anyway to distribute things better across multiple GPUs? are there any more lightweight versions of the base models i could use to get this running?

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions