Model card and the original license

README.md

@@ -27,6 +27,8 @@ for what was measured).
 |---|---|---|---|---|
 | bf16 | [`bf16/config.json`](https://huggingface.co/moonbox-dev/DeepFilterNet3-MLX/blob/main/bf16/config.json) | 0.9 kB | `cc96fe1bf688` | below threshold (experimental): SI-SDR 8.58 dB, 0.66 dB below the original on one clip; DNSMOS 2.93; output vs the original's 25.2 / 16.3 dB |
 | bf16 | [`bf16/model.safetensors`](https://huggingface.co/moonbox-dev/DeepFilterNet3-MLX/blob/main/bf16/model.safetensors) | 4.3 MB | `ad9d67c69a00` | below threshold (experimental): SI-SDR 8.58 dB, 0.66 dB below the original on one clip; DNSMOS 2.93; output vs the original's 25.2 / 16.3 dB |
+| fp16 | [`fp16/config.json`](https://huggingface.co/moonbox-dev/DeepFilterNet3-MLX/blob/main/fp16/config.json) | 0.9 kB | `cc96fe1bf688` | SI-SDR 8.62 dB (original 8.71), DNSMOS 2.92 (2.95); output vs the original's 25.3 dB mean / 17.7 dB worst clip (mlx-audio float32: 25.3 / 17.7) |
+| fp16 | [`fp16/model.safetensors`](https://huggingface.co/moonbox-dev/DeepFilterNet3-MLX/blob/main/fp16/model.safetensors) | 4.3 MB | `a0e5b4f0c96a` | SI-SDR 8.62 dB (original 8.71), DNSMOS 2.92 (2.95); output vs the original's 25.3 dB mean / 17.7 dB worst clip (mlx-audio float32: 25.3 / 17.7) |
 
 ## How to run