Instructions to use dgrauet/void-model-mlx-q8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use dgrauet/void-model-mlx-q8 with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] huggingface-cli download --local-dir void-model-mlx-q8 dgrauet/void-model-mlx-q8
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
Add int8 weights (group_size 64), config, quantize_config — repo was empty (#dogfood 2026-07-20)
Browse files- config.json +10 -0
- quantize_config.json +6 -0
- void_pass1.safetensors +3 -0
- void_pass2.safetensors +3 -0
config.json
ADDED
|
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_type": "void-transformer",
|
| 3 |
+
"source": "netflix-void",
|
| 4 |
+
"architecture": "CogVideoXTransformer3DModel",
|
| 5 |
+
"passes": ["void_pass1", "void_pass2"],
|
| 6 |
+
"notes": {
|
| 7 |
+
"patch_embed": "Linear with in_dim=384 (in_channels=48 [16 latent + 16 VAE-mask + 16 VAE-video] * patch_volume=8).",
|
| 8 |
+
"base_model": "Uses VAE and T5 from CogVideoX-Fun-V1.5-5b-InP."
|
| 9 |
+
}
|
| 10 |
+
}
|
quantize_config.json
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"quantization": {
|
| 3 |
+
"bits": 8,
|
| 4 |
+
"group_size": 64
|
| 5 |
+
}
|
| 6 |
+
}
|
void_pass1.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:760732a3ca55159ccdbaba1e0756af0d64954fdb0b799d7918562ad2a7882889
|
| 3 |
+
size 6684031105
|
void_pass2.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:77f756785bc77ce779ca6edfb16f6ed525383b13616d9228cdddc497c5a258c0
|
| 3 |
+
size 6684031105
|