[Core][Spec Decode] Replace EAGLE group heuristics with explicit draft-layer tags (#61455)

Signed-off-by: Dakai An <dakaian108@gmail.com>
Signed-off-by: Yifan Qiao <yifanqiao@inferact.ai>
Co-authored-by: Yifan Qiao <yifanqiao@inferact.ai>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
Dakai An
2026-10-17 08:22:33 +00:00
committed by GitHub
co-authored by Yifan Qiao Claude Opus 5.5
parent 08a65e287e
commit c3de0b37e2
10 changed files with 160 additions and 354 deletions
+3 -8
View File
@@ -103,15 +103,8 @@ def test_mtp_load_model_unified(mock_get_model, mock_get_layers, mock_get_pp_gro
target_attn_layers = {"target_attn_1": mock.MagicMock()}
all_attn_layers = {**target_attn_layers, "draft_attn_1": mock.MagicMock()}
target_indexer_layers: dict = {}
all_indexer_layers: dict = {}
mock_get_layers.side_effect = [
target_attn_layers,
target_indexer_layers,
all_attn_layers,
all_indexer_layers,
]
mock_get_layers.side_effect = [target_attn_layers, all_attn_layers]
mock_pp_group = mock.MagicMock()
mock_pp_group.world_size = 1
@@ -138,6 +131,8 @@ def test_mtp_load_model_unified(mock_get_model, mock_get_layers, mock_get_pp_gro
assert proposer.model.lm_head == target_model.lm_head
# MTP shares embed_tokens with target model
assert proposer.model.model.embed_tokens == target_model.model.embed_tokens
assert proposer._draft_attn_layer_names == {"draft_attn_1"}
assert all_attn_layers["draft_attn_1"].is_draft_layer is True
@pytest.mark.parametrize("num_speculative_tokens", [1])