refactor(moss_tts_nano): migrate to Pipeline + Deploy schema

Follow-up to #2958. MOSS-TTS-Nano was authored before the schema refactor
and still shipped as a legacy stage_configs yaml; this aligns it with the
same layout the 5 migrated TTS models now use.

- vllm_omni/model_executor/models/moss_tts_nano/pipeline.py declares
  MOSS_TTS_NANO_PIPELINE (single LLM_AR stage, owns_tokenizer, audio
  output, stop_token_ids=[2] as a hard EOS backstop).
- vllm_omni/deploy/moss_tts_nano.yaml holds runtime knobs (max_num_seqs,
  gpu_memory_utilization, enforce_eager, default_sampling_params,
  skip_mm_profiling); trust_remote_code stays at deploy top-level.
- vllm_omni/config/pipeline_registry.py registers the entry so the lazy
  registry can resolve it.
- moss_tts_nano/__init__.py exports MossTTSNanoForGeneration (VoxCPM2 pattern).
- Removed vllm_omni/model_executor/stage_configs/moss_tts_nano.yaml.

Examples, shell scripts, READMEs, and buildkite-invoked tests are
updated to use `vllm serve <model> --omni` / `--deploy-config`
(auto-load kicks in; no --stage-configs-path or --trust-remote-code).

Signed-off-by: Yueqian Lin <linyueqian@outlook.com>
This commit is contained in:
Yueqian Lin committed 2026-04-21 18:07:50 -04:00
1 parent 26bbe7b527
commit b03a41a7f9
12 files changed
+104 -101

No files matched your search

@@ -1,2 +1,5 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
from .modeling_moss_tts_nano import MossTTSNanoForGeneration
__all__ = ["MossTTSNanoForGeneration"]
@@ -0,0 +1,40 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
"""MOSS-TTS-Nano pipeline topology (frozen).
Single-stage AR TTS: text -> speech waveform in one pass. The 0.1B AR LM
and the MOSS-Audio-Tokenizer-Nano codec both run inside
``MossTTSNanoForGeneration.forward()``, which uses the VoxCPM-style
generator pattern (``inference_stream()`` stored per-request; one audio
chunk yielded per forward) to drive progressive streaming through the
AR scheduler.
"""
from vllm_omni.config.stage_config import (
PipelineConfig,
StageExecutionType,
StagePipelineConfig,
)
MOSS_TTS_NANO_PIPELINE = PipelineConfig(
model_type="moss_tts_nano",
model_arch="MossTTSNanoForCausalLM",
stages=(
StagePipelineConfig(
stage_id=0,
model_stage="moss_tts_nano",
execution_type=StageExecutionType.LLM_AR,
input_sources=(),
final_output=True,
final_output_type="audio",
owns_tokenizer=True,
engine_output_type="audio",
sampling_constraints={
"detokenize": False,
# compute_logits() forces EOS (token id 2) when the last
# streaming chunk is yielded; keep a hard backstop here.
"stop_token_ids": [2],
},
),
),
)
@@ -1,61 +0,0 @@
# Stage config for MOSS-TTS-Nano (OpenMOSS-Team/MOSS-TTS-Nano).
#
# Single AR stage: the 0.1B AR LM and MOSS-Audio-Tokenizer-Nano codec
# both run inside MossTTSNanoForGeneration.forward(). The model uses the
# VoxCPM-style generator pattern for streaming: inference_stream() is
# stored per-request, and each forward() call yields one audio chunk.
# compute_logits() controls EOS to keep requests alive until done.
#
# Verified on: 1x GPU with ≥2 GB VRAM, or CPU (slower).
# Output: 48 kHz, 2-channel stereo audio.
#
# Request additional_information fields:
# text (str) – text to synthesize [required]
# voice (str) – built-in voice preset, default "Junhao"
# mode (str) – "continuation" (default) | "voice_clone"
# prompt_audio_path (str) – path to reference audio for custom voice clone
# prompt_text (str) – reference transcript (continuation mode)
# max_new_frames (int) – max AR frames, default 375 (~14 s)
# text_temperature (float) – default 1.0
# audio_temperature (float) – default 0.8
# audio_top_k (int) – default 25
# audio_top_p (float) – default 0.95
# seed (int) – optional RNG seed
async_chunk: false
stage_args:
- stage_id: 0
stage_type: llm
is_comprehension: true
runtime:
devices: "0"
engine_args:
model_stage: moss_tts_nano
model_arch: MossTTSNanoForCausalLM
worker_type: ar
scheduler_cls: vllm_omni.core.sched.omni_ar_scheduler.OmniARScheduler
trust_remote_code: true
enforce_eager: true
engine_output_type: audio
gpu_memory_utilization: 0.3
enable_prefix_caching: false
skip_mm_profiling: true
max_num_seqs: 4
max_model_len: 4096
max_num_batched_tokens: 4096
final_output: true
final_output_type: audio
default_sampling_params:
temperature: 1.0
top_p: 1.0
top_k: 50
max_tokens: 4096
seed: 42
detokenize: false
runtime:
enabled: true
defaults:
window_size: -1
max_inflight: 4