-
Notifications
You must be signed in to change notification settings - Fork 1.6k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
OOM Issue #1923
Comments
#1894 check this one too |
Try running nvidia-smi to see which processes are using up your gpu memory. |
not sure why but my vram usage does not seem to imply that gpu_memory_utilization is being respected - i also have issues running a 7B model on an A40 (40GB vram) despite having no issues in using the HF backend (~28gb vram utilized there). specifically: it OOMs even before running any requests. Have verified that there is nothing else running on the GPU. However, passing |
@johnwee1 when you pass |
@johnwee1 confirmed on Discord that when using |
Hi! I am running evaluations but keep getting OOM errors. Here is my script:
I am using a 80GB A100. I have already decreased the
gpu_memory_utilization
andmax_model_len
, but the problem persists. I have tried Llama3-8b with the same hyperparameters and everything was just fine. Can anyone please tell me why this happens and how I can solve it? Thanks!The text was updated successfully, but these errors were encountered: