Skip to content

Added recipe for Qwen3-235b#145

Open
ankitkumar-quad wants to merge 6 commits intoAI-Hypercomputer:mainfrom
ankitkumar-quad:main
Open

Added recipe for Qwen3-235b#145
ankitkumar-quad wants to merge 6 commits intoAI-Hypercomputer:mainfrom
ankitkumar-quad:main

Conversation

@ankitkumar-quad
Copy link
Copy Markdown

No description provided.


```

### 5. Quantize the model using FP8
Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

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

Add an addditional optional section (5.1 Optional) to use a pre-quantized model, such as Nvidia's official FP8 checkpoint: https://huggingface.co/nvidia/Qwen3-235B-A22B-FP8

TP_SIZE=1
PP_SIZE=1
EP_SIZE=1

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

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

Add the setting for NCCL_P2P_LEVEL and a link to the doc for setting: https://docs.cloud.google.com/compute/docs/accelerator-optimized-machines#g4-gpu-p2p

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.

3 participants