[Spec Decode] Move SpecDecodeBaseProposer out of eagle.py (#40732)

Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
This commit is contained in:
Matthew Bonanni
2026-04-23 22:28:27 +00:00
committed by GitHub
parent 4a6dd1c3cc
commit cde8d24710
7 changed files with 1792 additions and 1785 deletions
+3 -3
View File
@@ -741,9 +741,9 @@ def test_set_inputs_first_pass_parallel_drafting():
@pytest.mark.parametrize("pp_size", [1, 2])
@pytest.mark.parametrize("use_distinct_embed_tokens", [True, False])
@pytest.mark.parametrize("use_distinct_lm_head", [True, False])
@mock.patch("vllm.v1.spec_decode.eagle.get_pp_group")
@mock.patch("vllm.v1.spec_decode.eagle.get_layers_from_vllm_config")
@mock.patch("vllm.v1.spec_decode.eagle.get_model")
@mock.patch("vllm.v1.spec_decode.llm_base_proposer.get_pp_group")
@mock.patch("vllm.v1.spec_decode.llm_base_proposer.get_layers_from_vllm_config")
@mock.patch("vllm.v1.spec_decode.llm_base_proposer.get_model")
def test_load_model(
mock_get_model,
mock_get_layers,
+3 -3
View File
@@ -61,9 +61,9 @@ def _create_mtp_proposer(num_speculative_tokens: int) -> EagleProposer:
return EagleProposer(vllm_config=vllm_config, device=DEVICE_TYPE)
@mock.patch("vllm.v1.spec_decode.eagle.get_pp_group")
@mock.patch("vllm.v1.spec_decode.eagle.get_layers_from_vllm_config")
@mock.patch("vllm.v1.spec_decode.eagle.get_model")
@mock.patch("vllm.v1.spec_decode.llm_base_proposer.get_pp_group")
@mock.patch("vllm.v1.spec_decode.llm_base_proposer.get_layers_from_vllm_config")
@mock.patch("vllm.v1.spec_decode.llm_base_proposer.get_model")
def test_mtp_load_model_unified(mock_get_model, mock_get_layers, mock_get_pp_group):
"""Test MTP-specific model loading with unified model approach."""