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 16, 10:45 UTC
[Misc] Remove user-facing error for removed VLM args (#9104)
Cyrus Leung committed
Oct 6, 2024 at 16:33 UTC
f22619fe96c842ee2406638678d2b60009d8ff14
2 files changed
+1
-13
docs/source/models/vlm.rst
-4
@@ -23,10 +23,6 @@ The :class:`~vllm.LLM` class can be instantiated in much the same way as languag
23
24
llm = LLM(model="llava-hf/llava-1.5-7b-hf")
25
26
-.. note::
27
- We have removed all vision language related CLI args in the ``0.5.1`` release. **This is a breaking change**, so please update your code to follow
28
- the above snippet. Specifically, ``image_feature_size`` can no longer be specified as we now calculate that internally for each model.
29
-
26
To pass an image to the model, note the following in :class:`vllm.inputs.PromptType`:
27
28
* ``prompt``: The prompt should follow the format that is documented on HuggingFace.
vllm/entrypoints/llm.py
+1
-9
@@ -180,15 +180,7 @@ class LLM:
180
181
if "disable_log_stats" not in kwargs:
182
kwargs["disable_log_stats"] = True
183
- removed_vision_keys = (
184
- "image_token_id",
185
- "image_feature_size",
186
- "image_input_shape",
187
- "image_input_type",
188
- )
189
- if any(k in kwargs for k in removed_vision_keys):
190
- raise TypeError(
191
- "There is no need to pass vision-related arguments anymore.")
183
+
184
engine_args = EngineArgs(
185
model=model,
186
tokenizer=tokenizer,