xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-06-03 18:09:07 +08:00
Code Issues Packages Projects Releases Wiki Activity

removed get_allowed_mm_limits() from models

Browse Source
This commit is contained in:
Anexdeus 2025-12-20 21:56:07 +03:00
parent 2b03137fca
commit 86c6c5cf00
3 changed files with 0 additions and 51 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

17
vllm/model_executor/models/qwen2_5_vl.py
View File

@ -1587,20 +1587,3 @@ class Qwen2_5_VLForConditionalGeneration(
vision_config = hf_config.vision_config vision_config = hf_config.vision_config
merge_size = vision_config.spatial_merge_size merge_size = vision_config.spatial_merge_size
return num_vision_tokens // merge_size**2 return num_vision_tokens // merge_size**2
def get_allowed_mm_limits(self) -> Mapping[str, int]:
"""Return the maximum allowed number of items for each modality."""
supported_mm_limits = self.get_supported_mm_limits()
mm_config = self.ctx.get_mm_config()
allowed_limits = dict[str, int]()
for modality, supported_limit in supported_mm_limits.items():
user_limit = mm_config.get_limit_per_prompt(modality)
allowed_limits[modality] = (
user_limit
if supported_limit is None
else min(user_limit, supported_limit)
)
return allowed_limits

17
vllm/model_executor/models/qwen2_vl.py
View File

@ -1491,23 +1491,6 @@ class Qwen2VLForConditionalGeneration(
tower_model="visual.", tower_model="visual.",
) )
def get_allowed_mm_limits(self) -> Mapping[str, int]:
"""Return the maximum allowed number of items for each modality."""
supported_mm_limits = self.get_supported_mm_limits()
mm_config = self.ctx.get_mm_config()
allowed_limits = dict[str, int]()
for modality, supported_limit in supported_mm_limits.items():
user_limit = mm_config.get_limit_per_prompt(modality)
allowed_limits[modality] = (
user_limit
if supported_limit is None
else min(user_limit, supported_limit)
)
return allowed_limits
def get_num_mm_encoder_tokens( def get_num_mm_encoder_tokens(
self, self,
num_image_tokens: int, num_image_tokens: int,

17
vllm/model_executor/models/qwen3_vl.py
View File

@ -2110,20 +2110,3 @@ class Qwen3VLForConditionalGeneration(
vision_config = hf_config.vision_config vision_config = hf_config.vision_config
merge_size = vision_config.spatial_merge_size merge_size = vision_config.spatial_merge_size
return num_vision_tokens // merge_size**2 return num_vision_tokens // merge_size**2
def get_allowed_mm_limits(self) -> Mapping[str, int]:
"""Return the maximum allowed number of items for each modality."""
supported_mm_limits = self.get_supported_mm_limits()
mm_config = self.ctx.get_mm_config()
allowed_limits = dict[str, int]()
for modality, supported_limit in supported_mm_limits.items():
user_limit = mm_config.get_limit_per_prompt(modality)
allowed_limits[modality] = (
user_limit
if supported_limit is None
else min(user_limit, supported_limit)
)
return allowed_limits