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

[Refactor] MLP weight prefetch to consistency with MoE Model's prefetching in terms of code and usage (#6442)

Browse Source 
### What this PR does / why we need it?
Refactor MLP weight prefetch to consistency with MoE Model's prefetching
in terms of code and usage.
Environments VLLM_ASCEND_ENABLE_PREFETCH_MLP,
VLLM_ASCEND_MLP_DOWN_PREFETCH_SIZE and
VLLM_ASCEND_MLP_GATE_UP_PREFETCH_SIZE is removed, usage as following:

--additional-config '{"weight_prefetch_config": { "enabled": true,
"prefetch_ratio": {"mlp": { "gate_up": 1.0, "down": 1.0} }}}'

### Does this PR introduce _any_ user-facing change?

### How was this patch tested?

- vLLM version: v0.14.1
- vLLM main:
dc917cceb8

---------

Signed-off-by: leo-pony <nengjunma@outlook.com>
This commit is contained in:
Nengjun Ma
2026-02-04 09:08:18 +08:00
committed by GitHub
parent fa56abea9f
commit 78fad4e348
18 changed files with 250 additions and 171 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
vllm_ascend/ascend_forward_context.py
View File

@@ -119,18 +119,8 @@ def set_ascend_forward_context(
if has_layer_idx(model_instance):
forward_context.layer_idx = model_instance.model.start_layer
# TODO(rjg-lyh): refactor mlp weight prefetch method
# set for mlp weight prefetch
prefetch_mlp_enabled = (
envs_ascend.VLLM_ASCEND_ENABLE_PREFETCH_MLP
and forward_context.layer_idx is not None
and num_tokens is not None
and num_tokens < 500
)
if prefetch_mlp_enabled:
forward_context.prefetch_mlp_gate_up_proj = False
forward_context.prefetch_mlp_down_proj = False
forward_context.prefetch_mlp_enabled = prefetch_mlp_enabled
forward_context.prefetch_mlp_gate_up_proj = False
forward_context.prefetch_mlp_down_proj = False
forward_context.model_instance = model_instance
forward_context.is_draft_model = is_draft_model