Skip to content
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

[MODEL] Add qwen moe support #24

Merged
merged 25 commits into from
Jun 20, 2024
Merged

[MODEL] Add qwen moe support #24

merged 25 commits into from
Jun 20, 2024

Conversation

LRL-ModelCloud
Copy link
Collaborator

No description provided.

@Qubitium
Copy link
Collaborator

Qubitium commented Jun 20, 2024

@bozheng-hit We have refractored your code to be more generic so that the same dynamic_expert_index variable expansion code can be re-used by other expert models. Currently we are quantizing and doing validation to make sure ppl before and after quant is in normal range. We should have this merged in the next couple of hours.

@Qubitium Qubitium changed the title Add qwen moe support [MODEL] Add qwen moe support Jun 20, 2024
@Qubitium
Copy link
Collaborator

Qubitium commented Jun 20, 2024

TEST PASSED:

Native PPL: 6.4653
Quant PPL: 6.8459

Note we only a rudimentary quant for testing and not a full-quality quant. The PPL values here is only for regression/sanity check.

@Qubitium Qubitium merged commit 95b2fe6 into main Jun 20, 2024
1 of 2 checks passed
@Qubitium Qubitium deleted the add-qwen-moe-support branch June 20, 2024 16:47
DeJoker pushed a commit to DeJoker/GPTQModel that referenced this pull request Jul 19, 2024
DeJoker pushed a commit to DeJoker/GPTQModel that referenced this pull request Jul 19, 2024
* support Qwen2MoE

* add description

* change order

* add type and type hint

* fix wrong args name and order

* fix need Qwen2MoeGPTQ

* update README.md

* add require_true_sequential property

* add dynamic_expert_layer_index property

* add todo

* shorten name

* use getattr()

* reduce torch/triton requirements to torch/triton 2.0.0

* rename inside_layer_modules to layer_modules

* print layers log

* Update perplexity.py

* Update model.py

* rename

* remove unused log

---------

Co-authored-by: bozheng-hit <dsoul0621@gmail.com>
Co-authored-by: Bo Zheng <368586905@qq.com>
Co-authored-by: LRL-ModelCloud <lrl@modelcloud.ai>
Co-authored-by: ZX-ModelCloud <zx@modelcloud.ai>
Co-authored-by: diegomontoya <xing@fictionpress.com>
Co-authored-by: Qubitium-ModelCloud <qubitium@modelcloud.ai>
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.

4 participants