Skip to content

feat(quant): move quantization cookbook helpers to framework - #233

Merged
pengcuo merged 4 commits into
NVIDIA:mainfrom
mutichung:mutic/cosmos_fp8_quantization_cookbook
Sep 4, 2026
Merged

pengcuo merged 4 commits into
NVIDIA:mainfrom
mutichung:mutic/cosmos_fp8_quantization_cookbook

Conversation

@mutichung

Copy link
Copy Markdown
Contributor

Moves all python helpers under src/ from the quantization cookbooks in NVIDIA/cosmos#298 into cosmos-framework. The cookbooks will become thin wrappers that calls the functions to quantize checkpoints.

Blocks NVIDIA/cosmos#298.

Signed-off-by: Mu-Ti Chung <mutic@nvidia.com>
Signed-off-by: Mu-Ti Chung <mutic@nvidia.com>
@pengcuo
pengcuo merged commit 0fe093e into NVIDIA:main Sep 4, 2026
9 checks passed
qizhe pushed a commit to SimplexSim/cosmos-framework that referenced this pull request Oct 8, 2026
)

Moves all python helpers under `src/` from the quantization cookbooks in
NVIDIA/cosmos#298 into `cosmos-framework`. The
cookbooks will become thin wrappers that calls the functions to quantize
checkpoints.

Blocks NVIDIA/cosmos#298.

---------

Signed-off-by: Mu-Ti Chung <mutic@nvidia.com>
Co-authored-by: lfengad <liangf@nvidia.com>
Co-authored-by: pengcuo <pzeren@nvidia.com>
@mutichung
mutichung deleted the mutic/cosmos_fp8_quantization_cookbook branch October 15, 2026 13:30
foreverlms pushed a commit that referenced this pull request Oct 20, 2026
## Summary

Following #233, add Cosmos3-Edge support for FP8 quantization.

## Changes

* Add Edge support to quantization pipeline.
* Update generation tower name.
* Apply `skip_filter` during quantization.
* Fuse qkv projection temporarily during quantization.

---------

Signed-off-by: Mu-Ti Chung <mutic@nvidia.com>
Co-authored-by: pengcuo <pzeren@nvidia.com>
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