-
Notifications
You must be signed in to change notification settings - Fork 254
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
CUDA_ERROR_ILLEGAL_ADDRESS when running Llama3 and Llama3.1 #783
Comments
@ShelbyJenkins can you reproduce the issue if you ensure running without PagedAttention? |
Sorry for the delay. Been updating things on my backend. I just upgraded to the newest hash. Love the new API <3 Yes, PagedAttention should be disabled based on how I init it right? Additionally, I'm feeding it a mapper, so it should disable it by default.
|
Interestingly, phi3.5 works with my setup. Mistral Nemo and Llama3.2 3b do have the CUDA_ERROR however. |
@ShelbyJenkins I'll take a look at what is causing this. |
This occurs when using two GPUs, but it does not occur when I use just the one.
I made sure to update to the docker image used in the dockerfile.
commit: a702c6d (from earlier this week)
Tried the latest commit from today and using Llama3_1_8bInstruct.
Originally posted by @ShelbyJenkins in #651 (comment)
The text was updated successfully, but these errors were encountered: