OpenMOSS / OpenMOSS/MOSS-TTS-Nano
Working with CUDA
Open
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 4.4k
- Forks
- 555
- PR merge metrics
- No merged PRs in 30d
Description
what?
Research on Inference Acceleration for MOSS-TTS-Nano on GPUs
Usage
.venv\Scripts\python.exe app_onnx.py --execution-provider cuda
Acceleration Results
| Method | RTF (Real-Time Factor) | Relative Speedup |
|---|---|---|
| PyTorch CUDA (Baseline) | ~1.9x | 1.0x |
| ONNX CUDA (Recommended) | ~1.34x | 1.4x faster |
| PyTorch + nq=8 | ~1.5x | 1.3x faster |
Discarded Options
- CUDA Graph: The ONNX graph contains CPU↔GPU Memcpy nodes, preventing enablement (errors occurred during testing);
- torch.compile: Autoregressive dynamic shapes; offers limited gains and carries compatibility risks;
- flash-attn: The GTX 1080 Ti is based on the Pascal architecture (sm_61) and lacks Tensor Cores; unsupported;
Contributor guide
No contributing guide indexed for this repository
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start with app_onnx.py and run the documented CUDA execution-provider command in the issue. Check that the reported PyTorch and ONNX acceleration results can be reproduced, then document the required setup and confirmed results so users can follow the CUDA workflow.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python, pytorch
- Domain
- cli, machine-learning, performance
- Issue type
- Documentation
- Difficulty
- 3/5
- Estimated time
- 1-2 days
- Activity status
- Quiet
- Clarity
- Needs clarification
- Newbie friendliness
- 45/100