Instructions to use nvidia/personaplex-7b-v1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Moshi
How to use nvidia/personaplex-7b-v1 with Moshi:
# pip install moshi # Run the interactive web server python -m moshi.server --hf-repo "nvidia/personaplex-7b-v1" # Then open https://localhost:8998 in your browser
# pip install moshi import torch from moshi.models import loaders # Load checkpoint info from HuggingFace checkpoint = loaders.CheckpointInfo.from_hf_repo("nvidia/personaplex-7b-v1") # Load the Mimi audio codec mimi = checkpoint.get_mimi(device="cuda") mimi.set_num_codebooks(8) # Encode audio (24kHz, mono) wav = torch.randn(1, 1, 24000 * 10) # [batch, channels, samples] with torch.no_grad(): codes = mimi.encode(wav.cuda()) decoded = mimi.decode(codes) - Notebooks
- Google Colab
- Kaggle
[Community Fixes] Python 3.13 Support & Memory Optimization PRs on GitHub
Hi everyone!
I've submitted two pull requests on GitHub to address common issues discussed here:
PR #48 - Python 3.13 Support
Fixes: Issue #22 - Installation fails on Python 3.13
Problem: The pyproject.toml had torch >= 2.2.0, < 2.5 which blocked Python 3.13 users (PyTorch added 3.13 wheels in v2.5.0).
Solution: Relaxed dependency constraints while maintaining backward compatibility:
torch >= 2.2.0(removed upper bound)numpy >= 1.26, < 3safetensors >= 0.4.0- And more...
PR Link: https://github.com/NVIDIA/personaplex/pull/48
PR #49 - Remove Dead Code (~200MB GPU Memory Savings)
Fixes: Issue #46 - Unused other_mimi instance
Problem: There was a duplicate MimiModel instance (other_mimi) that:
- Consumed ~200MB GPU memory
- Ran encode/decode on every frame but results were discarded
- Was never actually used
Solution: Removed all other_mimi references from server.py and offline.py.
PR Link: https://github.com/NVIDIA/personaplex/pull/49
These fixes should help users running PersonaPlex on:
- Python 3.13 environments
- Consumer GPUs with limited VRAM (every 200MB helps!)
Feel free to test the fixes by installing from my fork:
pip install git+https://github.com/ThanhNguyxn/personaplex.git@fix/python313-compatibility#subdirectory=moshi
Feedback welcome!