Upload folder using huggingface_hub
Browse files
audio_dit/config.json
ADDED
|
@@ -0,0 +1,27 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_class_name": "WanAudioModel",
|
| 3 |
+
"_diffusers_version": "0.36.0",
|
| 4 |
+
"_name_or_path": "checkpoints/MOVA-720p/audio_dit",
|
| 5 |
+
"add_control_adapter": false,
|
| 6 |
+
"dim": 1536,
|
| 7 |
+
"eps": 1e-06,
|
| 8 |
+
"ffn_dim": 8960,
|
| 9 |
+
"freq_dim": 256,
|
| 10 |
+
"fuse_vae_embedding_in_latents": false,
|
| 11 |
+
"has_image_input": false,
|
| 12 |
+
"has_image_pos_emb": false,
|
| 13 |
+
"has_ref_conv": false,
|
| 14 |
+
"in_dim": 128,
|
| 15 |
+
"in_dim_control_adapter": 24,
|
| 16 |
+
"num_heads": 12,
|
| 17 |
+
"num_layers": 30,
|
| 18 |
+
"out_dim": 128,
|
| 19 |
+
"patch_size": [
|
| 20 |
+
1
|
| 21 |
+
],
|
| 22 |
+
"require_clip_embedding": true,
|
| 23 |
+
"require_vae_embedding": true,
|
| 24 |
+
"seperated_timestep": false,
|
| 25 |
+
"text_dim": 4096,
|
| 26 |
+
"vae_type": "dac"
|
| 27 |
+
}
|
audio_dit/diffusion_pytorch_model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:06804e8aba4e75477fa922e92ad0b03ae2ea2cade7e8829dd36d76fb5c3827d1
|
| 3 |
+
size 2838470672
|