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

Refactor Transformers backend to use mixins (#26906)

Browse Source 
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2025-10-16 22:50:39 +01:00
committed by GitHub
parent b2f78cbad4
commit fb5e10d3fb
17 changed files with 1510 additions and 1248 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
tests/models/registry.py
View File

@@ -912,11 +912,11 @@ _TRANSFORMERS_BACKEND_MODELS = {
"TransformersForCausalLM": _HfExamplesInfo(
"hmellor/Ilama-3.2-1B", trust_remote_code=True
),
"TransformersForMultimodalLM": _HfExamplesInfo("BAAI/Emu3-Chat-hf"),
"TransformersMultiModalForCausalLM": _HfExamplesInfo("BAAI/Emu3-Chat-hf"),
"TransformersMoEForCausalLM": _HfExamplesInfo(
"allenai/OLMoE-1B-7B-0924", min_transformers_version="4.57.0.dev0"
),
"TransformersMoEForMultimodalLM": _HfExamplesInfo(
"TransformersMultiModalMoEForCausalLM": _HfExamplesInfo(
"Qwen/Qwen3-VL-30B-A3B-Instruct", min_transformers_version="4.57.0.dev0"
),
"TransformersMoEEmbeddingModel": _HfExamplesInfo(
@@ -925,6 +925,10 @@ _TRANSFORMERS_BACKEND_MODELS = {
"TransformersMoEForSequenceClassification": _HfExamplesInfo(
"Qwen/Qwen3-30B-A3B", min_transformers_version="4.57.0.dev0"
),
"TransformersMultiModalEmbeddingModel": _HfExamplesInfo("google/gemma-3-4b-it"),
"TransformersMultiModalForSequenceClassification": _HfExamplesInfo(
"google/gemma-3-4b-it"
),
}
_EXAMPLE_MODELS = {

2
tests/models/test_initialization.py
View File

@@ -37,7 +37,7 @@ MINIMAL_MODEL_ARCH_LIST = [
"JinaVLForRanking",
"InternVLChatModel",
"InternLM2ForRewardModel",
"TransformersForMultimodalLM",
"TransformersMultiModalForCausalLM",
"PrithviGeoSpatialMAE",
"UltravoxModel",
"DeepSeekMTPModel",

6
tests/models/test_transformers.py
View File

@@ -211,11 +211,7 @@ def test_embed_loading(vllm_runner, model):
def test_pooling(hf_runner, vllm_runner, example_prompts, arch):
model = get_model(arch)
vllm_kwargs = dict(
max_model_len=None,
model_impl="transformers",
compilation_config=dict(cudagraph_capture_sizes=[8]),
)
vllm_kwargs = dict(max_model_len=None, model_impl="transformers")
hf_kwargs = dict()
if arch == "TransformersEmbeddingModel":