[Model] Extend device-side mm normalization to Qwen3VL/Qwen3.5/Qwen4Next (#51289)
Signed-off-by: cjackal <44624812+cjackal@users.noreply.github.com> Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
This commit is contained in:
co-authored by
mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
parent
07f1ba2ec6
commit
491f44adfa
@@ -14,6 +14,7 @@ from vllm.distributed import get_pp_group
|
||||
from vllm.model_executor.layers.fused_moe.utils import (
|
||||
is_model_fused_shared_expert_compatible,
|
||||
)
|
||||
from vllm.model_executor.layers.fusion.mm_input_norm import build_mm_input_norm
|
||||
from vllm.model_executor.layers.logits_processor import LogitsProcessor
|
||||
from vllm.model_executor.layers.mamba.gdn.qwen_gdn_linear_attn import (
|
||||
QwenGatedDeltaNetAttention,
|
||||
@@ -897,6 +898,7 @@ class Qwen4ExpForConditionalGeneration(
|
||||
config.vision_config,
|
||||
norm_eps=config.text_config.rms_norm_eps,
|
||||
quant_config=quant_config,
|
||||
input_norm=build_mm_input_norm(self.model_config),
|
||||
prefix=maybe_prefix(prefix, "visual"),
|
||||
)
|
||||
|
||||
|
||||
@@ -13,6 +13,7 @@ from vllm.distributed import get_pp_group
|
||||
from vllm.model_executor.layers.fused_moe.utils import (
|
||||
is_model_fused_shared_expert_compatible,
|
||||
)
|
||||
from vllm.model_executor.layers.fusion.mm_input_norm import build_mm_input_norm
|
||||
from vllm.model_executor.layers.logits_processor import LogitsProcessor
|
||||
from vllm.model_executor.layers.mamba.gdn.qwen_gdn_linear_attn import (
|
||||
QwenGatedDeltaNetAttention,
|
||||
@@ -939,6 +940,7 @@ class Qwen4ExpForConditionalGeneration(
|
||||
config.vision_config,
|
||||
norm_eps=config.text_config.rms_norm_eps,
|
||||
quant_config=quant_config,
|
||||
input_norm=build_mm_input_norm(self.model_config),
|
||||
prefix=maybe_prefix(prefix, "visual"),
|
||||
)
|
||||
|
||||
|
||||
Reference in New Issue
Block a user