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

[Bugfix] Fix disaggregated pd error (#2242)

Browse Source 
### What this PR does / why we need it?
Fix `ascend_env has no attr VLLM_ASCEND_ENABLE_CHUNK_MC2`, remove
useless lines

- vLLM version: v0.10.0
- vLLM main:
9edd1db02b

---------

Signed-off-by: wangli <wangli858794774@gmail.com>
This commit is contained in:
Li Wang
2025-08-06 19:48:10 +08:00
committed by GitHub
parent 26fc36b0e0
commit 57b9f02185
1 changed files with 0 additions and 16 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
vllm_ascend/worker/model_runner_v1.py
View File

@@ -646,22 +646,6 @@ class NPUModelRunner(LoRAModelRunnerMixin):
return maybe_padded_num_tokens, None, with_prefill, enable_dbo
return num_tokens, None, with_prefill, enable_dbo
if self.is_kv_producer and not envs_ascend.VLLM_ASCEND_ENABLE_CHUNK_MC2:
num_tokens_across_dp = torch.tensor([num_tokens] * self.dp_size,
device="cpu",
dtype=torch.int32)
return num_tokens, num_tokens_across_dp, True, enable_dbo
if self.is_kv_consumer and self.torchair_graph_enabled and len(
self.torchair_graph_batch_sizes
) == 1 and not self.in_profile_run:
max_num_decode_tokens = self.torchair_graph_batch_sizes[0]
num_tokens_across_dp = torch.tensor([max_num_decode_tokens] *
self.dp_size,
device="cpu",
dtype=torch.int32)
return max_num_decode_tokens, num_tokens_across_dp, False, enable_dbo
maybe_padded_num_tokens = num_tokens
num_tokens_across_dp, with_prefill, enable_dbo = self._get_forward_metadata_across_dp(
num_tokens, with_prefill, enable_dbo)