Text-to-Speech
MLX
Chinese
English
auk-base
speech
speech-editing
apple-silicon
audio
8bit
quantized
Instructions to use vanch007/AuK-Base-MLX-8bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use vanch007/AuK-Base-MLX-8bit with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] huggingface-cli download --local-dir AuK-Base-MLX-8bit vanch007/AuK-Base-MLX-8bit
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
Upload Native MLX 8-Bit quantized AuK-Base model weights
Browse files- README.md +30 -0
- config.json +14 -0
- dit_8bit.safetensors +3 -0
- vae.safetensors +3 -0
README.md
ADDED
|
@@ -0,0 +1,30 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language:
|
| 3 |
+
- zh
|
| 4 |
+
- en
|
| 5 |
+
tags:
|
| 6 |
+
- speech
|
| 7 |
+
- text-to-speech
|
| 8 |
+
- speech-editing
|
| 9 |
+
- mlx
|
| 10 |
+
- apple-silicon
|
| 11 |
+
- audio
|
| 12 |
+
- 8bit
|
| 13 |
+
- quantized
|
| 14 |
+
license: mit
|
| 15 |
+
pipeline_tag: text-to-speech
|
| 16 |
+
---
|
| 17 |
+
|
| 18 |
+
# 馃殌 AuK-Base-MLX-8bit (Apple Silicon 8-Bit Quantized)
|
| 19 |
+
|
| 20 |
+
<p align="center">
|
| 21 |
+
<a href="https://github.com/vanch007/mlx-AuK"><b>馃敆 GitHub Source & Web Demo: vanch007/mlx-AuK</b></a> 路
|
| 22 |
+
<a href="https://huggingface.co/vanch007/AuK-Base-MLX"><b>Full Precision MLX Weights</b></a>
|
| 23 |
+
</p>
|
| 24 |
+
|
| 25 |
+
This repository provides the **8-Bit affine quantized MLX port** of **AuK Base**, Tencent Hunyuan's 1.5B speech foundation model.
|
| 26 |
+
|
| 27 |
+
- **Main GitHub Repository**: [https://github.com/vanch007/mlx-AuK](https://github.com/vanch007/mlx-AuK)
|
| 28 |
+
- **Original Project**: [Tencent-Hunyuan/AuK](https://github.com/Tencent-Hunyuan/AuK)
|
| 29 |
+
- **Quantization**: 8-bit group-wise affine quantization ()
|
| 30 |
+
- **Weight Size**: **0.56 GB** (90.1% reduction from 5.70 GB)
|
config.json
ADDED
|
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_type": "auk-base",
|
| 3 |
+
"is_flash": false,
|
| 4 |
+
"dim": 1536,
|
| 5 |
+
"heads": 24,
|
| 6 |
+
"num_layers": 10,
|
| 7 |
+
"num_single_layers": 20,
|
| 8 |
+
"latent_dim": 64,
|
| 9 |
+
"sample_rate": 24000,
|
| 10 |
+
"downsample_rate": 480,
|
| 11 |
+
"default_nfe": 32,
|
| 12 |
+
"default_cfg": 2.0,
|
| 13 |
+
"sway_sampling_coef": -1.0
|
| 14 |
+
}
|
dit_8bit.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:962a1420495225da13a885f560b11e4168b460c6e4e2af8d0618116e1a40bf48
|
| 3 |
+
size 597322493
|
vae.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0f5857fd0d6b916d161c73fef433e0ccb42484a2dad3aa08e8a51975e29a0645
|
| 3 |
+
size 637322604
|