Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
base_model: Wan-AI/Wan2.2-T2V-A14B
|
| 4 |
+
library_name: mlx
|
| 5 |
+
pipeline_tag: text-to-video
|
| 6 |
+
tags:
|
| 7 |
+
- mlx
|
| 8 |
+
- apple-silicon
|
| 9 |
+
- wan2.2
|
| 10 |
+
- text-to-video
|
| 11 |
+
- video-generation
|
| 12 |
+
- mixture-of-experts
|
| 13 |
+
---
|
| 14 |
+
|
| 15 |
+
# Wan2.2-T2V-A14B — MLX (bf16)
|
| 16 |
+
|
| 17 |
+
Native **MLX** (Apple Silicon) conversion of [Wan-AI/Wan2.2-T2V-A14B](https://huggingface.co/Wan-AI/Wan2.2-T2V-A14B), packaged as a turnkey, self-contained snapshot for the [SceneWorks](https://huggingface.co/SceneWorks) app.
|
| 18 |
+
|
| 19 |
+
Wan2.2 A14B is a **high/low-noise mixture-of-experts** text-to-video model — two transformers switched at the noise boundary.
|
| 20 |
+
|
| 21 |
+
## Contents (self-contained, bf16)
|
| 22 |
+
| file | what |
|
| 23 |
+
|---|---|
|
| 24 |
+
| `high_noise_model.safetensors` | high-noise expert DiT (~28.6 GB) |
|
| 25 |
+
| `low_noise_model.safetensors` | low-noise expert DiT (~28.6 GB) |
|
| 26 |
+
| `t5_encoder.safetensors` | UMT5-XXL text encoder (~11.4 GB) |
|
| 27 |
+
| `vae.safetensors` | Wan z16 VAE |
|
| 28 |
+
| `tokenizer.json` | UMT5 tokenizer |
|
| 29 |
+
| `config.json` | architecture config |
|
| 30 |
+
|
| 31 |
+
Quantization (Q4/Q8) is applied **at load** by the engine — these weights are full bf16.
|
| 32 |
+
|
| 33 |
+
## Provenance
|
| 34 |
+
- **Source:** `Wan-AI/Wan2.2-T2V-A14B` (Apache-2.0).
|
| 35 |
+
- **Converted with:** the SceneWorks native Rust MLX converter (`mlx-gen-wan`, converter id `wan_t2v_14b`), dtype `bfloat16`, dense (no baked-in quant).
|
| 36 |
+
- Lean snapshot — only the files the MLX engine loads.
|
| 37 |
+
|
| 38 |
+
## License
|
| 39 |
+
**Apache-2.0**, inherited from the upstream model. This repository redistributes a converted copy of the upstream Apache-2.0 weights, with attribution, as permitted by that license. See the [source model card](https://huggingface.co/Wan-AI/Wan2.2-T2V-A14B).
|