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 14, 05:30 UTC
[CI] enable idefics3 and fuyu-8b test in multimodal test (#23790)
Signed-off-by: zjy0516 <riverclouds.zhu@qq.com>
Jiangyun Zhu committed
Aug 28, 2025 at 14:51 UTC
c07a73317d202c2dad67f12893fcddb6d3664950
1 file changed
+15
-21
tests/models/multimodal/generation/test_common.py
+15
-21
@@ -189,23 +189,21 @@ VLM_TEST_SETTINGS = {
189
},
190
marks=[pytest.mark.core_model],
191
),
192
- # FIXME(Isotr0py): Enable this test after
193
- # https://github.com/huggingface/transformers/pull/39470 released
194
- # "idefics3-transformers": VLMTestInfo(
195
- # models=["HuggingFaceTB/SmolVLM-256M-Instruct"],
196
- # test_type=(VLMTestType.IMAGE, VLMTestType.MULTI_IMAGE),
197
- # prompt_formatter=lambda img_prompt:f"<|begin_of_text|>User:{img_prompt}<end_of_utterance>\nAssistant:", # noqa: E501
198
- # img_idx_to_prompt=lambda idx: "<image>",
199
- # max_model_len=8192,
200
- # max_num_seqs=2,
201
- # auto_cls=AutoModelForImageTextToText,
202
- # hf_output_post_proc=model_utils.idefics3_trunc_hf_output,
203
- # image_size_factors=[(0.25, 0.5, 1.0)],
204
- # vllm_runner_kwargs={
205
- # "model_impl": "transformers",
206
- # },
207
- # marks=[pytest.mark.core_model],
208
- # ),
192
+ "idefics3-transformers": VLMTestInfo(
193
+ models=["HuggingFaceTB/SmolVLM-256M-Instruct"],
194
+ test_type=(VLMTestType.IMAGE, VLMTestType.MULTI_IMAGE),
195
+ prompt_formatter=lambda img_prompt:f"<|begin_of_text|>User:{img_prompt}<end_of_utterance>\nAssistant:", # noqa: E501
196
+ img_idx_to_prompt=lambda idx: "<image>",
197
+ max_model_len=8192,
198
+ max_num_seqs=2,
199
+ auto_cls=AutoModelForImageTextToText,
200
+ hf_output_post_proc=model_utils.idefics3_trunc_hf_output,
201
+ image_size_factors=[(0.25, 0.5, 1.0)],
202
+ vllm_runner_kwargs={
203
+ "model_impl": "transformers",
204
+ },
205
+ marks=[pytest.mark.core_model],
206
+ ),
207
# Pixel values from processor are not 4D or 5D arrays
208
"qwen2_5_vl-transformers": VLMTestInfo(
209
models=["Qwen/Qwen2.5-VL-3B-Instruct"],
@@ -322,10 +320,6 @@ VLM_TEST_SETTINGS = {
320
vllm_output_post_proc=model_utils.fuyu_vllm_to_hf_output,
321
num_logprobs=10,
322
image_size_factors=[(), (0.25,), (0.25, 0.25, 0.25), (0.25, 0.2, 0.15)],
325
- # FIXME(Isotr0py): This model is broken in Transformers v4.54.1, we
326
- # should enable this again after the fix is released:
327
- # https://github.com/huggingface/transformers/pull/39915
328
- marks=[pytest.mark.skip("HF model is broken")],
323
),
324
"gemma3": VLMTestInfo(
325
models=["google/gemma-3-4b-it"],