feat(audio): add ACE-Step 1.5 music generation support - #5413
Merged
Conversation
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Code Review
This pull request integrates the ACE-Step 1.5 text-to-music model into Xinference, adding the necessary model specifications, documentation, and a vendored copy of the upstream ACE-Step 1.5 runtime. The review feedback focuses on improving multi-GPU environment compatibility, specifically by ensuring proper device indexing for VRAM checks, setting the active CUDA device, and refining device comparison logic. Additionally, suggestions were made to prevent potential AttributeError crashes through defensive fallback patterns and nested attribute checks, log warnings during fallback environment variable parsing, and add the Telugu language code 'te' to the codespell allowlist.
qinxuye
reviewed
Aug 23, 2026
qinxuye
reviewed
Aug 23, 2026
Keep model-level index strategy for PyTorch engines so uv can resolve ARM torchao from PyPI alongside CUDA wheels.
Bundle selected inference sources and nano-vllm so launches no longer require Git package installs. Exclude training, UI, API, tests, examples, and unused DiT variants.
Use the selected CUDA device for VRAM checks and device validation. Harden optional adapter, config, and environment handling.
Minamiyama
force-pushed
the
models/ace-step1.5
branch
from
August 24, 2026 13:00
3911049 to
4051920
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Add built-in ACE-Step 1.5 support for text-to-music generation.
Changes
ACE-Step1.5as an audio model using the PyTorch engine.ACE-Step/Ace-Step1.5, revisionmainACE-Step/Ace-Step1.5, revisionmasteracestep-v15-turboDiT and official VAE.acestep-5Hz-lm-1.7Bwithpt,vllm, andmlxbackends.inputinstructVendored runtime
Vendor the inference-only ACE-Step 1.5 runtime from upstream commit:
14c0211d5a0653b0f63e27686f4c3f151b4d8629The vendored source includes the required ACE-Step inference components and
bundled
nano-vllmruntime. Training code, datasets, API servers, CLI/GradioUI, tests, examples, development metadata, and unused DiT variants are
excluded.
Runtime dependencies such as PyTorch, Transformers, Diffusers, and TorchAO
remain managed by the Xinference model virtual environment. Model launch no
longer requires installing ACE-Step source directly from GitHub.
The upstream MIT licenses and source revision information are retained.
Current scope
This initial integration supports the components included in the complete
ACE-Step 1.5 checkpoint bundle:
acestep-v15-turboacestep-5Hz-lm-1.7BArbitrary standalone DiT, LM, and VAE combinations are left for follow-up work.