xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-09 18:54:55 +08:00
Code Issues Packages Projects Releases Wiki Activity

[ROCm] Remove unnecessary assertion of max_model_len in ROCM_AITER_MLA attention backend. (#18938)

Browse Source 
Signed-off-by: vllmellm <vllm.ellm@embeddedllm.com>
This commit is contained in:
vllmellm 2025-05-30 13:33:17 +08:00 committed by GitHub
parent 5acf828d99
commit 77b6e74fe2
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 0 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
vllm/attention/backends/rocm_aiter_mla.py
View File

@ -132,8 +132,6 @@ class AiterMLAMetadataBuilder(MLACommonMetadataBuilder[AiterMLAMetadata]):
def __init__(self, input_builder: "ModelInputForGPUBuilder"):
super().__init__(input_builder)
assert self.runner.model_config.max_model_len == 32768,\
"AITER MLA requires max model len to be set to 32768"
assert self.block_size == 1, "AITER MLA requires only block size 1."
def prepare(self):

3
vllm/v1/attention/backends/mla/rocm_aiter_mla.py
View File

@ -66,9 +66,6 @@ class AiterMLAMetadataBuilder(MLACommonMetadataBuilder[AiterMLAMetadata]):
def __init__(self, runner, kv_cache_spec: AttentionSpec,
block_table: BlockTable):
super().__init__(runner, kv_cache_spec, block_table)
max_model_len = self.runner.model_config.max_model_len
assert max_model_len == 32768,\
"AITER MLA requires max_model_len=32768"
assert self.kv_cache_spec.block_size == 1, "AITER MLA" \
"only supports block size 1."