[Model] Add YuE2-3B text-to-music (model + /v1/audio/speech adapter) (#7886)

Signed-off-by: Youyuan Li <105433145+Darcy-Lee@users.noreply.github.com>
Signed-off-by: princepride <wangzhipeng628@gmail.com>
Co-authored-by: princepride <wangzhipeng628@gmail.com>
This commit is contained in:
Youyuan Liandprincepride authored and GitHub committed 2026-10-01 18:04:31 +08:00
1 parent a038b38179
commit 423f34326e
27 files changed
+3294 -6

No files matched your search

@@ -483,6 +483,12 @@ _OMNI_MODELS = {
"acoustic",
"MiniMaxMusic3AcousticForConditionalGeneration",
),
## YuE2-3B (text-to-music; single-stage native-AR with terminal NAR/VAE)
"Yue2ForCausalLM": (
"yue2",
"yue2",
"Yue2ForCausalLM",
),
}
@@ -0,0 +1,7 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM-Omni project
"""YuE2-3B text-to-music model package."""
from .pipeline import YUE2_PIPELINE
__all__ = ["YUE2_PIPELINE"]
@@ -0,0 +1,53 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM-Omni project
"""YuE2-3B pipeline topology.
Single-stage native-AR text-to-music (gepard pattern). The whole
Mixture-of-Transformers checkpoint — AR path, NAR path, projection heads —
lives in one ``model.safetensors``, so both the autoregressive sampling and
the terminal flow-matching + VAE pass run inside one stage and no weights are
duplicated across stages. The VAE decoder is a separate repository
(``m-a-p/YuE2-Vae``), loaded in ``load_weights`` from ``$YUE2_VAE`` or the
default hub id.
Sampling is model-owned. ``sampling_constraints`` pins ``detokenize`` and the
union of both phase end tokens; ``extra_args`` carry the phase and the
request-local sampling preset, and must be present in the stage's
``default_sampling_params`` so ``has_sampling_extra_args`` turns on and the
per-request args reach the model's forward.
"""
from vllm_omni.config.stage_config import (
PipelineConfig,
StageExecutionType,
StagePipelineConfig,
)
from .yue2 import STOP_TOKEN_IDS
YUE2_PIPELINE = PipelineConfig(
model_type="yue2",
default_deploy_config_name="yue2.yaml",
model_arch="Yue2ForCausalLM",
stages=(
StagePipelineConfig(
stage_id=0,
model_stage="yue2",
execution_type=StageExecutionType.LLM_AR,
input_sources=(),
final_output=True,
final_output_type="audio",
owns_tokenizer=True,
engine_output_type="audio",
# Whole-song audio ships once, on the finishing step; the audio
# accumulator concatenates over time and a single tensor just
# passes through.
sampling_constraints={
"detokenize": False,
"stop_token_ids": list(STOP_TOKEN_IDS),
},
),
),
)
__all__ = ["YUE2_PIPELINE"]
File diff suppressed because it is too large. Load diff