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

[refactor] refactor deepseek-related files (#2849)

Browse Source 
### What this PR does / why we need it?
This PR deletes ~2K lines of code about deepseek modeling. It falls back
CustomDeepseekV2 modules to original vllm implementations and adapts
some modifications in vllm about deepseek and moe.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
E2E  vllm serving with torchair graph mode and eager mode.

- vLLM version: v0.10.2
- vLLM main:
759ef49b15

---------

Signed-off-by: linfeng-yuan <1102311262@qq.com>
Signed-off-by: Yizhou Liu <liu_yizhou@outlook.com>
Co-authored-by: yiz-liu <136800916+yiz-liu@users.noreply.github.com>
Co-authored-by: Yizhou Liu <liu_yizhou@outlook.com>
This commit is contained in:
linfeng-yuan
2025-09-16 14:13:07 +08:00
committed by GitHub
parent 18ca7861f6
commit 1c5900327b
18 changed files with 295 additions and 1899 deletions
Download Patch File Download Diff File Expand all files Collapse all files

27
vllm_ascend/models/__init__.py
View File

@@ -4,10 +4,6 @@ import vllm_ascend.envs as envs_ascend
def register_model():
ModelRegistry.register_model(
"DeepSeekMTPModel",
"vllm_ascend.models.deepseek_mtp:CustomDeepSeekMTP")
ModelRegistry.register_model(
"Qwen2VLForConditionalGeneration",
"vllm_ascend.models.qwen2_vl:AscendQwen2VLForConditionalGeneration")
@@ -23,22 +19,17 @@ def register_model():
"vllm_ascend.models.qwen2_5_vl_without_padding:AscendQwen2_5_VLForConditionalGeneration_Without_Padding"
)
if envs_ascend.VLLM_ASCEND_ENABLE_DBO:
ModelRegistry.register_model(
"DeepseekV2ForCausalLM",
"vllm_ascend.models.deepseek_dbo:CustomDeepseekDBOForCausalLM")
ModelRegistry.register_model(
"DeepseekV2ForCausalLM",
"vllm_ascend.models.deepseek_v2:CustomDeepseekV2ForCausalLM")
ModelRegistry.register_model(
"DeepseekV3ForCausalLM",
"vllm_ascend.models.deepseek_dbo:CustomDeepseekDBOForCausalLM")
else:
ModelRegistry.register_model(
"DeepseekV2ForCausalLM",
"vllm_ascend.models.deepseek_v2:CustomDeepseekV2ForCausalLM")
ModelRegistry.register_model(
"DeepseekV3ForCausalLM",
"vllm_ascend.models.deepseek_v3:CustomDeepseekV3ForCausalLM")
ModelRegistry.register_model(
"DeepseekV3ForCausalLM",
"vllm_ascend.models.deepseek_v3:CustomDeepseekV3ForCausalLM")
ModelRegistry.register_model(
"DeepSeekMTPModel",
"vllm_ascend.models.deepseek_mtp:CustomDeepSeekMTP")
ModelRegistry.register_model(
"Qwen3MoeForCausalLM",