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

[Lint]Style: Convert vllm-ascend/ to ruff format(Batch #6) (#6001)

Browse Source 
### What this PR does / why we need it?
| File Path |
| :--- |
| ` vllm_ascend/eplb/adaptor/abstract_adaptor.py` |
| ` vllm_ascend/eplb/adaptor/vllm_adaptor.py` |
| ` vllm_ascend/eplb/core/eplb_device_transfer_loader.py` |
| ` vllm_ascend/eplb/core/eplb_utils.py` |
| ` vllm_ascend/eplb/core/eplb_worker.py` |
| ` vllm_ascend/eplb/core/policy/policy_abstract.py` |
| ` vllm_ascend/eplb/core/policy/policy_default_eplb.py` |
| ` vllm_ascend/eplb/core/policy/policy_factory.py` |
| ` vllm_ascend/eplb/core/policy/policy_flashlb.py` |
| ` vllm_ascend/eplb/core/policy/policy_random.py` |
| ` vllm_ascend/eplb/core/policy/policy_swift_balancer.py` |
| ` vllm_ascend/eplb/eplb_updator.py` |
| ` vllm_ascend/eplb/utils.py` |
| ` vllm_ascend/model_loader/netloader/executor/elastic_load.py` |
| ` vllm_ascend/model_loader/netloader/executor/netloader_pg.py` |
| ` vllm_ascend/model_loader/netloader/interaction/elastic.py` |
| ` vllm_ascend/model_loader/netloader/load.py` |
| ` vllm_ascend/model_loader/netloader/netloader.py` |
| ` vllm_ascend/model_loader/netloader/utils.py` |
| ` vllm_ascend/patch/platform/__init__.py` |
| ` vllm_ascend/patch/platform/patch_balance_schedule.py` |
| ` vllm_ascend/patch/platform/patch_ec_connector.py` |
| ` vllm_ascend/patch/platform/patch_mamba_config.py` |
| ` vllm_ascend/patch/platform/patch_multiproc_executor.py` |
| ` vllm_ascend/patch/platform/patch_sched_yield.py` |


- vLLM version: v0.13.0
- vLLM main:
2c24bc6996

---------

Signed-off-by: MrZ20 <2609716663@qq.com>
This commit is contained in:
SILONG ZENG
2026-01-24 22:08:33 +08:00
committed by GitHub
parent 153da1a669
commit 4e53c1d900
26 changed files with 894 additions and 1148 deletions
Download Patch File Download Diff File Expand all files Collapse all files

30
vllm_ascend/patch/platform/patch_mamba_config.py
View File

@@ -38,7 +38,8 @@ def verify_and_update_config(cls, vllm_config) -> None:
block_size=1,
num_kv_heads=model_config.get_num_kv_heads(parallel_config),
head_size=model_config.get_head_size(),
dtype=kv_cache_dtype).page_size_bytes
dtype=kv_cache_dtype,
).page_size_bytes
model_cls, _ = ModelRegistry.resolve_model_cls(
model_config.architecture,
@@ -58,23 +59,20 @@ def verify_and_update_config(cls, vllm_config) -> None:
# block size to multiple of 16, so let's suggest a value
# that would work (note: FA is currently not compatible
# with mamba layers, use FlashInfer instead).
attn_block_size = block_alignment_bytes * cdiv(
mamba_page_size, block_alignment_bytes * attn_page_size_1_token)
attn_block_size = block_alignment_bytes * cdiv(mamba_page_size, block_alignment_bytes * attn_page_size_1_token)
# override attention block size if either (a) the
# user has not set it or (b) the user has set it
# too small.
if (cache_config.block_size is None
or cache_config.block_size < attn_block_size):
if cache_config.block_size is None or cache_config.block_size < attn_block_size:
cache_config.block_size = attn_block_size
logger.info(
"Setting attention block size to %d tokens "
"to ensure that attention page size is >= mamba page size.",
attn_block_size)
"Setting attention block size to %d tokens to ensure that attention page size is >= mamba page size.",
attn_block_size,
)
# compute new attention page size
attn_page_size = \
cache_config.block_size * attn_page_size_1_token
attn_page_size = cache_config.block_size * attn_page_size_1_token
assert attn_page_size >= mamba_page_size
@@ -83,15 +81,15 @@ def verify_and_update_config(cls, vllm_config) -> None:
return
# pad mamba page size to exactly match attention
if (cache_config.mamba_page_size_padded is None
or cache_config.mamba_page_size_padded != attn_page_size):
cache_config.mamba_page_size_padded = (attn_page_size)
mamba_padding_pct = 100 * (attn_page_size -
mamba_page_size) / mamba_page_size
if cache_config.mamba_page_size_padded is None or cache_config.mamba_page_size_padded != attn_page_size:
cache_config.mamba_page_size_padded = attn_page_size
mamba_padding_pct = 100 * (attn_page_size - mamba_page_size) / mamba_page_size
logger.info(
"Padding mamba page size by %.2f%% to ensure "
"that mamba page size and attention page size are "
"exactly equal.", mamba_padding_pct)
"exactly equal.",
mamba_padding_pct,
)
vllm.model_executor.models.config.HybridAttentionMambaModelConfig.verify_and_update_config = verify_and_update_config