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

[main] rename device type (#5099)

Browse Source 
### What this PR does / why we need it?
Rename `_910B` to `A2`;
Rename `_910_93` to `A3`;
Rename `_910_95` to `A5`;

- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c

Signed-off-by: zzzzwwjj <1183291235@qq.com>
This commit is contained in:
zzzzwwjj
2025-12-17 14:08:19 +08:00
committed by GitHub
parent 4144376e88
commit 06b82e7503
16 changed files with 47 additions and 48 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
vllm_ascend/ascend_forward_context.py
View File

@@ -244,7 +244,7 @@ def select_moe_comm_method(num_tokens: int,
if not vllm_config.parallel_config.enable_expert_parallel:
moe_comm_type = MoECommType.ALLGATHER
elif soc_version in {AscendDeviceType._910B}:
elif soc_version in {AscendDeviceType.A2}:
if (num_tokens <= mc2_tokens_capacity
and vllm_config.parallel_config.world_size_across_dp /
vllm_config.parallel_config.pipeline_parallel_size >= 16):
@@ -256,7 +256,7 @@ def select_moe_comm_method(num_tokens: int,
else:
moe_comm_type = MoECommType.ALLGATHER
elif soc_version in {AscendDeviceType._910_93}:
elif soc_version in {AscendDeviceType.A3}:
ascend_config = get_ascend_config()
dynamic_eplb = ascend_config.dynamic_eplb or ascend_config.expert_map_record_path
# TODO: drop the EP-size guard when dispatch_ffn_combine supports larger EP sizes