Upload config.json with huggingface_hub
Browse files- config.json +40 -0
config.json
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_class_name": "FoleyOmniConfig",
|
| 3 |
+
"_name_or_path": "CocoBro/Foley-Omni",
|
| 4 |
+
"architectures": [
|
| 5 |
+
"FoleyOmniDiT"
|
| 6 |
+
],
|
| 7 |
+
"model_type": "t2a",
|
| 8 |
+
"patch_size": [1],
|
| 9 |
+
"dim": 3072,
|
| 10 |
+
"ffn_dim": 14336,
|
| 11 |
+
"freq_dim": 256,
|
| 12 |
+
"num_heads": 24,
|
| 13 |
+
"num_layers": 30,
|
| 14 |
+
"in_dim": 20,
|
| 15 |
+
"out_dim": 20,
|
| 16 |
+
"text_len": 512,
|
| 17 |
+
"window_size": [-1, -1],
|
| 18 |
+
"qk_norm": true,
|
| 19 |
+
"cross_attn_norm": true,
|
| 20 |
+
"eps": 1e-6,
|
| 21 |
+
"temporal_rope_scaling_factor": 0.19676,
|
| 22 |
+
"sync_drop_from_context": false,
|
| 23 |
+
"sync_add_to_audio": true,
|
| 24 |
+
"sync_add_to_audio_each_block": true,
|
| 25 |
+
"sync_use_adaln": false,
|
| 26 |
+
"sync_use_proj": true,
|
| 27 |
+
"sync_use_adapter": true,
|
| 28 |
+
"sync_adapter_type": "depthwise_conv",
|
| 29 |
+
"sync_adapter_kernel_size": 5,
|
| 30 |
+
"sync_adapter_expansion": 2.0,
|
| 31 |
+
"use_continuous_time_context_rope": false,
|
| 32 |
+
"continuous_time_context_rope_mode": "auto",
|
| 33 |
+
"use_text_context_rope": true,
|
| 34 |
+
"use_clip_context_rope": true,
|
| 35 |
+
"use_sync_context_rope": true,
|
| 36 |
+
"context_rope_theta": 10000.0,
|
| 37 |
+
"context_text_rope_scaling": 1.0,
|
| 38 |
+
"context_clip_rope_scaling": 1.0,
|
| 39 |
+
"context_sync_rope_scaling": 1.0
|
| 40 |
+
}
|