A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 16:00 UTC
[Misc] Remove redundant code (#16098)
Signed-off-by: chaunceyjiang <chaunceyjiang@gmail.com>
Chauncey committed
Apr 6, 2025 at 11:03 UTC
13affc432de047057dce9d47a9a9152bf4a7b50c
1 file changed
+1
-3
vllm/entrypoints/chat_utils.py
+1
-3
@@ -498,7 +498,7 @@ class BaseMultiModalItemTracker(ABC, Generic[_T]):
498
hf_config.image_token_index)
499
if model_type in ("aya_vision", "chameleon", "deepseek_vl_v2",
500
"internvl_chat", "skywork_chat", "NVLM_D",
501
- "h2ovl_chat"):
501
+ "h2ovl_chat", "idefics3"):
502
return "<image>"
503
if model_type == "mllama":
504
return "<|image|>"
@@ -506,8 +506,6 @@ class BaseMultiModalItemTracker(ABC, Generic[_T]):
506
return "<|vision_start|><|image_pad|><|vision_end|>"
507
if model_type == "molmo":
508
return ""
509
- if model_type == "idefics3":
510
- return "<image>"
509
if model_type == "aria":
510
return "<|fim_prefix|><|img|><|fim_suffix|>"
511
if model_type == "gemma3":