Upload Ref2VA/transformer/config.json
Browse files
Ref2VA/transformer/config.json
ADDED
|
@@ -0,0 +1,27 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_class_name": "MiniMaxH3DiTModel",
|
| 3 |
+
"_diffusers_version": "0.32.2",
|
| 4 |
+
"hidden_size": 5376,
|
| 5 |
+
"num_layers": 50,
|
| 6 |
+
"token_refiner_num_layers": 2,
|
| 7 |
+
"num_attention_heads": 56,
|
| 8 |
+
"attention_head_dim": 128,
|
| 9 |
+
"ffn_hidden_size": 14336,
|
| 10 |
+
"latents_dim": 24,
|
| 11 |
+
"audio_latents_dim": 32,
|
| 12 |
+
"patch_size": [
|
| 13 |
+
1,
|
| 14 |
+
2,
|
| 15 |
+
2
|
| 16 |
+
],
|
| 17 |
+
"text_dim": 5120,
|
| 18 |
+
"timestep_input_dim": 256,
|
| 19 |
+
"time_embed_hidden_size": 5376,
|
| 20 |
+
"time_embed_dim": 2688,
|
| 21 |
+
"adaln_out_features": 96768,
|
| 22 |
+
"final_adaln_out_features": 10752,
|
| 23 |
+
"rope_inv_freq_len": 16,
|
| 24 |
+
"norm_eps": 1e-05,
|
| 25 |
+
"qk_norm_eps": 1e-05,
|
| 26 |
+
"final_norm_eps": 1e-05
|
| 27 |
+
}
|