[Refactor] Move profiling methods to MM budget (#33559)
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
@@ -82,6 +82,7 @@ from vllm.model_executor.models.interfaces_base import (
|
||||
is_text_generation_model,
|
||||
)
|
||||
from vllm.multimodal import MULTIMODAL_REGISTRY
|
||||
from vllm.multimodal.budget import MultiModalBudget
|
||||
from vllm.multimodal.inputs import (
|
||||
BatchedTensorInputs,
|
||||
MultiModalKwargsItem,
|
||||
@@ -180,7 +181,6 @@ from vllm.v1.worker.workspace import lock_workspace
|
||||
|
||||
from .utils import (
|
||||
AttentionGroup,
|
||||
MultiModalBudget,
|
||||
add_kv_sharing_layers_to_kv_cache_groups,
|
||||
bind_kv_cache,
|
||||
sanity_check_mm_encoder_outputs,
|
||||
@@ -5104,7 +5104,7 @@ class GPUModelRunner(
|
||||
# modality with the max possible input tokens even when
|
||||
# it supports multiple.
|
||||
dummy_modality = mm_budget.get_modality_with_max_tokens()
|
||||
max_mm_items_per_batch = mm_budget.max_items_per_batch_by_modality[
|
||||
max_mm_items_per_batch = mm_budget.mm_max_items_per_batch[
|
||||
dummy_modality
|
||||
]
|
||||
|
||||
|
||||
Reference in New Issue
Block a user