Skip to content

Conversation

dtrifiro
Copy link

@dtrifiro dtrifiro commented May 9, 2025

This depends on https://github.com/neuralmagic/nm-cicd/pull/103 to allow for the accuracy/guidellm workflows to actually use the accelerator-specific overrides

  • mistralai/Mixtral-8x7B-Instruct-v0.1: add rocm accuracy server override (avoids using tensor-parallel=8)
  • Llama-3.1-8B-Instruct add accuracy/server-rocm (set gpu_memory_utilization to avoid OOM errors)

@@ -0,0 +1,4 @@
# https://huggingface.co/mistralai/Mixtral-8x7B-Instruct-v0.1
model: 'mistralai/Mixtral-8x7B-Instruct-v0.1'
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The model config here can be removed.

@dtrifiro dtrifiro force-pushed the rocm-configs branch 3 times, most recently from cec96c2 to bae1481 Compare May 12, 2025 16:21
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants