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

Upgrade to 0.11.1 newest vllm commit (#3762)

Browse Source 
### What this PR does / why we need it?

c9461e05a4

Fix ```spec decode rejection sampler```, caused by
https://github.com/vllm-project/vllm/pull/26060
Fix some ```import```, caused by
https://github.com/vllm-project/vllm/pull/27374
Fix ```scheduler_config.send_delta_data```, caused by
https://github.com/vllm-project/vllm-ascend/pull/3719
Fix ```init_with_cudagraph_sizes```, caused by
https://github.com/vllm-project/vllm/pull/26016
Fix ```vl model```of replacing PatchEmbed's conv3d to linear layer,
caused by https://github.com/vllm-project/vllm/pull/27418

### Does this PR introduce _any_ user-facing change?
N/A

### How was this patch tested?
CI passed with new added/existing test.


- vLLM version: v0.11.0rc3
- vLLM main:
c9461e05a4

---------

Signed-off-by: Icey <1790571317@qq.com>
This commit is contained in:
Icey
2025-10-28 14:55:03 +08:00
committed by GitHub
parent f846bd20e4
commit a7450db1bd
12 changed files with 175 additions and 51 deletions
Download Patch File Download Diff File Expand all files Collapse all files

45
vllm_ascend/utils.py
View File

@@ -311,6 +311,41 @@ def get_max_hidden_layers(hf_config) -> int:
return max(layer_counts)
# Update cudagraph capture sizes for vllm config
def update_cudagraph_capture_sizes(vllm_config: VllmConfig,
cudagraph_capture_sizes: List[int]):
valid_max_size = (cudagraph_capture_sizes[-1]
if cudagraph_capture_sizes else 0)
if (vllm_config.compilation_config.max_cudagraph_capture_size is not None
and vllm_config.compilation_config.max_cudagraph_capture_size
!= valid_max_size):
if vllm_config.compilation_config.cudagraph_capture_sizes is not None:
raise ValueError(
"customized max_cudagraph_capture_size"
f"(={vllm_config.compilation_config.max_cudagraph_capture_size}) "
"should be consistent with the max value of "
f"cudagraph_capture_sizes(={valid_max_size})")
logger.warning(
"Truncating max_cudagraph_capture_size to %d",
valid_max_size,
)
vllm_config.compilation_config.max_cudagraph_capture_size = valid_max_size
if vllm_config.compilation_config.cudagraph_capture_sizes is not None and len(
cudagraph_capture_sizes) < len(
vllm_config.compilation_config.cudagraph_capture_sizes):
logger.warning(
("cudagraph_capture_sizes specified in compilation_config"
" %s is overridden by config %s"),
vllm_config.compilation_config.cudagraph_capture_sizes,
cudagraph_capture_sizes,
)
vllm_config.compilation_config.cudagraph_capture_sizes = cudagraph_capture_sizes
vllm_config.compilation_config.post_init_cudagraph_sizes()
def update_aclgraph_sizes(vllm_config: VllmConfig) -> None:
"""Update ACL graph capture sizes based on hardware limitations"""
# NOTE: Currently, we can only capture 1800 graphs at most,
@@ -402,7 +437,10 @@ def update_aclgraph_sizes(vllm_config: VllmConfig) -> None:
indices[0], indices[-1] = 0, len(original_sizes) - 1
sampled_sizes = [original_sizes[i] for i in indices]
compilation_config.init_with_cudagraph_sizes(sampled_sizes)
if vllm_version_is("0.11.0"):
compilation_config.init_with_cudagraph_sizes(sampled_sizes)
else:
update_cudagraph_capture_sizes(vllm_config, sampled_sizes)
logger.info(
"Adjusted ACL graph batch sizes for %s model (layers: %d): %d → %d sizes",
@@ -433,7 +471,10 @@ def update_aclgraph_sizes(vllm_config: VllmConfig) -> None:
if original_sizes[0] < (num_speculative_tokens + 1) * max_num_seqs:
enlarged_sizes = [(num_speculative_tokens + 1) * size
for size in original_sizes]
compilation_config.init_with_cudagraph_sizes(enlarged_sizes)
if vllm_version_is("0.11.0"):
compilation_config.init_with_cudagraph_sizes(enlarged_sizes)
else:
update_cudagraph_capture_sizes(vllm_config, enlarged_sizes)
logger.info(
"Adjusted ACL graphs: %s → %s for speculative decoding",
original_sizes, enlarged_sizes)