EngineX/xc-llm-ascend
3
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

fix deepseek torchair precision (#3635)

Browse Source 
### What this PR does / why we need it?
The precision of deepseek torchair is broken by #3465 , which due to the origin patch or rmsnorm in torchair. This PR fixes the precision of deepseek torchair.

### Does this PR introduce _any_ user-facing change?
No

### How was this patch tested?
<!--
CI passed with new added/existing test.
If it was tested in a way different from regular unit tests, please
clarify how you tested step by step, ideally copy and paste-able, so
that other reviewers can test and check, and descendants can verify in
the future.
If tests were not added, please describe why they were not added and/or
why it was difficult to add.
-->

Signed-off-by: hust17yixuan <303660421@qq.com>
This commit is contained in:
Wang Yixuan
2025-10-22 20:20:32 +08:00
committed by GitHub
parent 984efdc0d0
commit edccd46d74
2 changed files with 34 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
vllm_ascend/torchair/utils.py
View File

@@ -229,10 +229,12 @@ def torchair_ops_patch():
AscendDeepseekScalingRotaryEmbedding.__init__ = deepseek_rope_init_func # type: ignore[method-assign]
AscendDeepseekScalingRotaryEmbedding.forward = native_rope_deepseek_forward # type: ignore[method-assign]
AscendRMSNorm.__init__ = torchair_layernorm.torchair_rmsnorm_init_ # type: ignore[method-assign]
AscendRMSNorm.forward_oot = torchair_layernorm.torchair_rmsnorm_forward_oot # type: ignore[method-assign]
AscendSiluAndMul.forward_oot = torchair_activation.torchair_silu_and_mul_forward_oot # type: ignore[method-assign]
AscendVocabParallelEmbedding.forward = vocab_embedding_forward # type: ignore[method-assign]
def super_kernel(prefix: str, option: str, enabled: bool = True):
return _super_kernel(prefix, option) if enabled else nullcontext()
return _super_kernel(prefix, option) if enabled else nullcontext()