@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / 1c768fe537
Fork 0

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, 05:45 UTC

[Doc] Explicitly state that InternVL 2.5 is supported (#10978)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>

Cyrus Leung committed Dec 8, 2024 at 00:58 UTC 1c768fe53713ef333d74a6645e6a59fb7516134f
3 files changed +4 -4
docs/source/models/supported_models.rst
+2 -2
@@ -547,9 +547,9 @@ Text Generation
547 - ✅︎
548 -
549 * - :code:`InternVLChatModel`
550 - - InternVL2
550 + - InternVL 2.5, Mono-InternVL, InternVL 2.0
551 - T + I\ :sup:`E+`
552 - - :code:`OpenGVLab/Mono-InternVL-2B`, :code:`OpenGVLab/InternVL2-4B`, :code:`OpenGVLab/InternVL2-8B`, etc.
552 + - :code:`OpenGVLab/InternVL2_5-4B`, :code:`OpenGVLab/Mono-InternVL-2B`, :code:`OpenGVLab/InternVL2-4B`, etc.
553 -
554 - ✅︎
555 * - :code:`LlavaForConditionalGeneration`
examples/offline_inference_vision_language.py
+1 -1
@@ -223,7 +223,7 @@ def run_internvl(question: str, modality: str):
223 # Stop tokens for InternVL
224 # models variants may have different stop tokens
225 # please refer to the model card for the correct "stop words":
226 - # https://huggingface.co/OpenGVLab/InternVL2-2B#service
226 + # https://huggingface.co/OpenGVLab/InternVL2-2B/blob/main/conversation.py
227 stop_tokens = ["<|endoftext|>", "<|im_start|>", "<|im_end|>", "<|end|>"]
228 stop_token_ids = [tokenizer.convert_tokens_to_ids(i) for i in stop_tokens]
229 return llm, prompt, stop_token_ids
examples/offline_inference_vision_language_multi_image.py
+1 -1
@@ -165,7 +165,7 @@ def load_internvl(question: str, image_urls: List[str]) -> ModelRequestData:
165 # Stop tokens for InternVL
166 # models variants may have different stop tokens
167 # please refer to the model card for the correct "stop words":
168 - # https://huggingface.co/OpenGVLab/InternVL2-2B#service
168 + # https://huggingface.co/OpenGVLab/InternVL2-2B/blob/main/conversation.py
169 stop_tokens = ["<|endoftext|>", "<|im_start|>", "<|im_end|>", "<|end|>"]
170 stop_token_ids = [tokenizer.convert_tokens_to_ids(i) for i in stop_tokens]
171