Upload config.json with huggingface_hub
Browse files- config.json +26 -0
config.json
ADDED
|
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_class_name": "DiT",
|
| 3 |
+
"_diffusers_version": "0.25.0",
|
| 4 |
+
"_name_or_path": "f5-tts/dit",
|
| 5 |
+
"act_fn": "silu",
|
| 6 |
+
"attention_head_dim": 64,
|
| 7 |
+
"cross_attention_dim": null,
|
| 8 |
+
"dropout": 0.1,
|
| 9 |
+
"ff_mult": 4,
|
| 10 |
+
"in_channels": 100,
|
| 11 |
+
"layers_per_block": 1,
|
| 12 |
+
"num_attention_heads": 8,
|
| 13 |
+
"num_hidden_layers": 8,
|
| 14 |
+
"out_channels": 100,
|
| 15 |
+
"sample_size": 2048,
|
| 16 |
+
"flip_sin_to_cos": false,
|
| 17 |
+
"freq_shift": 0,
|
| 18 |
+
"use_linear_projection": false,
|
| 19 |
+
"projection_class_embeddings_input_dim": null,
|
| 20 |
+
"num_classes": null,
|
| 21 |
+
"time_embedding_type": "fourier",
|
| 22 |
+
"timestep_post_act": "silu",
|
| 23 |
+
"time_cond_proj_dim": null,
|
| 24 |
+
"class_embed_type": null,
|
| 25 |
+
"transformer_in_channels": null
|
| 26 |
+
}
|