Skip to content

Conversation

jinyouzhi
Copy link

@jinyouzhi jinyouzhi commented Aug 19, 2025

Enable ERNIE 4.5 series text generation models including Dense and MoE.

This PR contains:

then validate on

@Wei-Lin-Intel

@czhu15
Copy link

czhu15 commented Aug 22, 2025

Did we do anything specific for Gaudi? and have you verified 2 models on Gaudi?

@jinyouzhi
Copy link
Author

Did we do anything specific for Gaudi? and have you verified 2 models on Gaudi?

Basically no modifications. Yes, already validated 2 models on Gaudi.

  • baidu/ERNIE-4.5-0.3B-PT
Tasks Version Filter n-shot Metric Value Stderr
gsm8k 3 flexible-extract 5 exact_match 0.2266 ± 0.0262
strict-match 5 exact_match 0.1406 ± 0.0218
  • baidu/ERNIE-4.5-21B-A3B-PT
Tasks Version Filter n-shot Metric Value Stderr
gsm8k 3 flexible-extract 5 exact_match 0.8594 ± 0.0218
strict-match 5 exact_match 0.8438 ± 0.0227

Copy link

@czhu15 czhu15 left a comment

Choose a reason for hiding this comment

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

LGTM

@czhu15
Copy link

czhu15 commented Aug 25, 2025

cc @yangulei, @ranzhejiang

@czhu15 czhu15 merged commit e502a28 into HabanaAI:aice/v1.22.0 Aug 25, 2025
1 check passed
czhu15 pushed a commit that referenced this pull request Sep 5, 2025
Enable ERNIE 4.5 series text generation models including Dense and MoE. 

This PR contains:
- cherry-pick vllm-project#20220
- bugfix vllm-project#21735 (name mistake)

then validate on

- [ERNIE-4.5-0.3B-PT](https://huggingface.co/baidu/ERNIE-4.5-0.3B-PT)
-
[ERNIE-4.5-21B-A3B-PT](https://huggingface.co/baidu/ERNIE-4.5-21B-A3B-PT)

@Wei-Lin-Intel

---------

Signed-off-by: wangyafeng <[email protected]>
Signed-off-by: vasqu <[email protected]>
Signed-off-by: DarkLight1337 <[email protected]>
Co-authored-by: CSWYF3634076 <[email protected]>
Co-authored-by: Anton Vlasjuk <[email protected]>
Co-authored-by: DarkLight1337 <[email protected]>
czhu15 pushed a commit that referenced this pull request Sep 5, 2025
Enable ERNIE 4.5 series text generation models including Dense and MoE. 

This PR contains:
- cherry-pick vllm-project#20220
- bugfix vllm-project#21735 (name mistake)

then validate on

- [ERNIE-4.5-0.3B-PT](https://huggingface.co/baidu/ERNIE-4.5-0.3B-PT)
-
[ERNIE-4.5-21B-A3B-PT](https://huggingface.co/baidu/ERNIE-4.5-21B-A3B-PT)

@Wei-Lin-Intel

---------

Signed-off-by: wangyafeng <[email protected]>
Signed-off-by: vasqu <[email protected]>
Signed-off-by: DarkLight1337 <[email protected]>
Co-authored-by: CSWYF3634076 <[email protected]>
Co-authored-by: Anton Vlasjuk <[email protected]>
Co-authored-by: DarkLight1337 <[email protected]>
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