[Misc]] Move processing context to multimodal directory (#25548)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2025-09-24 08:15:00 +00:00
committed by GitHub
parent 27ec3c78f3
commit 6488f3481b
13 changed files with 262 additions and 242 deletions
+9 -4
View File
@@ -11,8 +11,9 @@ import torch.nn.functional as F
from transformers import PretrainedConfig
from vllm.config import ModelConfig, ModelDType, RunnerOption
from vllm.inputs import InputContext
from vllm.logprobs import Logprob, PromptLogprobs, SampleLogprobs
from vllm.multimodal.processing import InputProcessingContext
from vllm.transformers_utils.tokenizer import cached_tokenizer_from_config
from .registry import HF_EXAMPLE_MODELS
@@ -264,7 +265,7 @@ def build_model_context(
limit_mm_per_prompt: Optional[dict[str, int]] = None,
mm_processor_cache_gb: int = 0,
):
"""Creates an InputContext for a given model.
"""Creates an InputProcessingContext for a given model.
Args:
model_id: ID of the model being considered.
@@ -273,7 +274,7 @@ def build_model_context(
limit_mm_per_prompt: Multimodal limits.
Returns:
InputContext for the model being considered.
InputProcessingContext for the model being considered.
"""
model_info = HF_EXAMPLE_MODELS.find_hf_info(model_id)
model_info.check_available_online(on_fail="skip")
@@ -298,7 +299,11 @@ def build_model_context(
enforce_eager=model_info.enforce_eager,
**model_config_kwargs,
)
return InputContext(model_config)
return InputProcessingContext(
model_config,
tokenizer=cached_tokenizer_from_config(model_config),
)
def check_embeddings_close(