FunCineForge MLX Qwen2-0.5B
MLX-converted Qwen2-0.5B backbone for FunCineForge TTS pipeline.
Performance
| Backend | tok/s | Speedup |
|---|---|---|
| PyTorch MPS (fp32) | 24.5 | 1.0x |
| MLX (fp32) | 116.6 | 4.76x |
Files
model.safetensors— MLX Qwen2-0.5B backbone weights (fp32)config.json— Model configurationcustom_weights.pt— FunCineForge custom layers (codec_embed, timespk_embed, codec_head, face_linear)tokenizer.json/tokenizer_config.json— Tokenizer files
Usage
Set use_mlx: true in decode.yaml. The model will be auto-downloaded on first run.
# decode.yaml
use_mlx: true
Requires: pip install mlx mlx-lm
- Downloads last month
- 45
Model size
0.5B params
Tensor type
F32
·
Hardware compatibility
Log In to add your hardware
Quantized
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support