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

[ModelRunner][Qwen3-Next] Fix attn_group initialization timing (#3477)

Browse Source 
### What this PR does / why we need it?
Fix attn_group initialization timing so that fix qwen3-next model

### Does this PR introduce _any_ user-facing change?

### How was this patch tested?

- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0

---------

Signed-off-by: MengqingCao <cmq0113@163.com>
This commit is contained in:
Mengqing Cao
2025-10-20 09:39:40 +08:00
committed by GitHub
parent 9e59fc1510
commit 6c65dd891f
1 changed files with 14 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
vllm_ascend/worker/model_runner_v1.py
View File

@@ -2694,6 +2694,9 @@ class NPUModelRunner(LoRAModelRunnerMixin):
""" """
kv_cache_config = deepcopy(kv_cache_config) kv_cache_config = deepcopy(kv_cache_config)
self.kv_cache_config = kv_cache_config self.kv_cache_config = kv_cache_config
self.may_add_encoder_only_layers_to_kv_cache_config()
# NOTE(cmq): initialize_attn_backend must before using self.attn_groups
self.initialize_attn_backend(kv_cache_config)
self.use_hybrid_blocks = (len(self.attn_groups) > 1) self.use_hybrid_blocks = (len(self.attn_groups) > 1)
# NOTE: Currently, we determine whether we need `num_accepted_tokens` through `MambaSpec`. # NOTE: Currently, we determine whether we need `num_accepted_tokens` through `MambaSpec`.
self.need_accepted_tokens = any([ self.need_accepted_tokens = any([
@@ -2702,8 +2705,6 @@ class NPUModelRunner(LoRAModelRunnerMixin):
]) ])
self.may_reinitialize_input_batch(kv_cache_config) self.may_reinitialize_input_batch(kv_cache_config)
self.may_add_encoder_only_layers_to_kv_cache_config()
self.initialize_attn_backend(kv_cache_config)
if self.use_sparse: if self.use_sparse:
kv_caches = self.initialize_kv_cache_tensors_deepseek_sfa( kv_caches = self.initialize_kv_cache_tensors_deepseek_sfa(
@@ -3100,6 +3101,8 @@ class NPUModelRunner(LoRAModelRunnerMixin):
block_sizes = [ block_sizes = [
kv_cache_group.kv_cache_spec.block_size kv_cache_group.kv_cache_spec.block_size
for kv_cache_group in kv_cache_config.kv_cache_groups for kv_cache_group in kv_cache_config.kv_cache_groups
if not isinstance(kv_cache_group.kv_cache_spec,
EncoderOnlyAttentionSpec)
] ]
# Generate kernel_block_sizes that matches each block_size # Generate kernel_block_sizes that matches each block_size
@@ -3109,7 +3112,11 @@ class NPUModelRunner(LoRAModelRunnerMixin):
kernel_block_sizes = [] kernel_block_sizes = []
for kv_cache_group_id, kv_cache_group in enumerate( for kv_cache_group_id, kv_cache_group in enumerate(
kv_cache_config.kv_cache_groups): kv_cache_config.kv_cache_groups):
if isinstance(kv_cache_group.kv_cache_spec, AttentionSpec):
if isinstance(kv_cache_group.kv_cache_spec,
EncoderOnlyAttentionSpec):
continue
elif isinstance(kv_cache_group.kv_cache_spec, AttentionSpec):
# This is an attention backend that supports virtual # This is an attention backend that supports virtual
# block splitting. Get the supported block sizes from # block splitting. Get the supported block sizes from
# the backend. # the backend.
@@ -3137,7 +3144,10 @@ class NPUModelRunner(LoRAModelRunnerMixin):
# of mamba block. In this case, BlockTable.block_size will never equal # of mamba block. In this case, BlockTable.block_size will never equal
# to kernel_block_sizes[0] # to kernel_block_sizes[0]
kernel_block_sizes.append([0]) kernel_block_sizes.append([0])
if kernel_block_sizes != [[self.cache_config.block_size]]:
if block_sizes != [
self.cache_config.block_size
] or kernel_block_sizes != [self.cache_config.block_size]:
assert self.cache_config.cpu_offload_gb == 0, ( assert self.cache_config.cpu_offload_gb == 0, (
"Cannot re-initialize the input batch when CPU weight " "Cannot re-initialize the input batch when CPU weight "
"offloading is enabled. See https://github.com/vllm-project/vllm/pull/18298 " # noqa: E501 "offloading is enabled. See https://github.com/vllm-project/vllm/pull/18298 " # noqa: E501