[V0 Deprecation] Remove V0 Spec Decode workers (#21152)

Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
This commit is contained in:
Woosuk Kwon
2025-07-18 21:47:50 -07:00
committed by GitHub
parent 9ffe905a41
commit dd572c0ab3
73 changed files with 191 additions and 14275 deletions
+4 -4
View File
@@ -457,12 +457,12 @@ _MULTIMODAL_EXAMPLE_MODELS = {
_SPECULATIVE_DECODING_EXAMPLE_MODELS = {
"EAGLEModel": _HfExamplesInfo("JackFram/llama-68m",
speculative_model="abhigoyal/vllm-eagle-llama-68m-random"), # noqa: E501
"MedusaModel": _HfExamplesInfo("JackFram/llama-68m",
speculative_model="abhigoyal/vllm-medusa-llama-68m-random"), # noqa: E501
"MLPSpeculatorPreTrainedModel": _HfExamplesInfo("JackFram/llama-160m",
speculative_model="ibm-ai-platform/llama-160m-accelerator"), # noqa: E501
# Temporarily disabled.
# TODO(woosuk): Re-enable this once the MLP Speculator is supported in V1.
# "MLPSpeculatorPreTrainedModel": _HfExamplesInfo("JackFram/llama-160m",
# speculative_model="ibm-ai-platform/llama-160m-accelerator"), # noqa: E501
"DeepSeekMTPModel": _HfExamplesInfo("luccafong/deepseek_mtp_main_random",
speculative_model="luccafong/deepseek_mtp_draft_random", # noqa: E501
trust_remote_code=True),
+9 -5
View File
@@ -72,11 +72,15 @@ def test_registry_model_property(model_arch, is_mm, init_cuda, is_ce):
@create_new_process_for_each_test()
@pytest.mark.parametrize("model_arch,is_pp,init_cuda", [
("MLPSpeculatorPreTrainedModel", False, False),
("DeepseekV2ForCausalLM", True, False),
("Qwen2VLForConditionalGeneration", True, True),
])
@pytest.mark.parametrize(
"model_arch,is_pp,init_cuda",
[
# TODO(woosuk): Re-enable this once the MLP Speculator is supported
# in V1.
# ("MLPSpeculatorPreTrainedModel", False, False),
("DeepseekV2ForCausalLM", True, False),
("Qwen2VLForConditionalGeneration", True, True),
])
def test_registry_is_pp(model_arch, is_pp, init_cuda):
assert ModelRegistry.is_pp_supported_model(model_arch) is is_pp