biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Multi Modal] Configurable MM Profiling (#25631)

Browse Source 
Signed-off-by: wwl2755 <wangwenlong2755@gmail.com>
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Wenlong Wang
2025-10-03 03:59:10 -07:00
committed by GitHub
parent 2ed3f20dba
commit 79aa244678
60 changed files with 654 additions and 99 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
tests/models/multimodal/processing/test_mllama4.py
View File

@@ -17,23 +17,23 @@ def test_profiling(model_id: str, max_model_len: int):
model_config_kwargs = {
"max_model_len": max_model_len,
}
mm_counts = {"image": 1}
ctx = build_model_context(
model_id,
model_config_kwargs=model_config_kwargs,
limit_mm_per_prompt={"image": 1},
limit_mm_per_prompt=mm_counts,
)
mm_config = ctx.get_mm_config()
processor = MULTIMODAL_REGISTRY.create_processor(ctx.model_config)
profiler = MultiModalProfiler(processor)
decoder_dummy_data = profiler.get_decoder_dummy_data(
max_model_len,
mm_counts=mm_config.limit_per_prompt,
mm_counts=mm_counts,
)
dummy_mm_data = processor.dummy_inputs.get_dummy_processor_inputs(
max_model_len,
mm_counts=mm_config.limit_per_prompt,
mm_counts=mm_counts,
)
hf_config = ctx.get_hf_config(Llama4Config)
@@ -58,7 +58,7 @@ def test_profiling(model_id: str, max_model_len: int):
profiled_tokens = profiler.get_mm_max_contiguous_tokens(
max_model_len,
mm_counts=mm_config.limit_per_prompt,
mm_counts=mm_counts,
)
assert total_tokens == profiled_tokens["image"]