RVC-Project/Retrieval-based-Voice-Conversion-WebUI

CUDA memory error, not using most of available memory

Open

#1,016 opened on Aug 15, 2023

View on GitHub
 (6 comments) (0 reactions) (0 assignees)Python (2,849 forks)batch import
help wanted

Repository metrics

Stars
 (18,427 stars)
PR merge metrics
 (No merged PRs in 30d)

Description

Hello there!

I get the CUDA out of memory error, but when I look at my GPU's memory usage, I see that it only uses up to 4GB of memory before crashing, while my GPU's dedicated memory is 8GB (my shared memory is also 8GB if that's somehow relevant). My GPU is 1070 GTX which is supported by CUDA and I have installed the latest drivers and all that.

Moreover, when I use cache it actually does work and uses up to 7GB without changing the batch size (I'm using 4). I am using WSL 2 for running RVC on my machine, and as far as I know, the WSL isn't limiting GPU memory usage, so theoretically it should use up to 100% of my 8GB of memories.

If someone might have a clue about what's going on and how can I fix it, I would be happy to know!

Contributor guide

Research direction
Investigate the PyTorch CUDA memory allocation settings. Try setting the environment variable PYTORCH CUDA ALLOC CONF to max split size mb:128 or similar. Check if WSL2 has any GPU memory limits. Monitor memory usage with nvidia smi.
Tech stack
pythonpytorch
Domain
backend
Issue type
Bug
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
3
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
Half day
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
Active
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
Clear
Prerequisites
PythonCUDA
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
65