Mirror text_encoder/config.json from cvssp/audioldm2@c8e7e189
Browse files
encoders/cvssp/audioldm2/text_encoder/config.json
ADDED
|
@@ -0,0 +1,35 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"ClapModel"
|
| 4 |
+
],
|
| 5 |
+
"audio_config": {
|
| 6 |
+
"depths": [
|
| 7 |
+
2,
|
| 8 |
+
2,
|
| 9 |
+
12,
|
| 10 |
+
2
|
| 11 |
+
],
|
| 12 |
+
"fusion_num_hidden_layers": 2,
|
| 13 |
+
"hidden_size": 1024,
|
| 14 |
+
"model_type": "clap_audio_model",
|
| 15 |
+
"patch_embeds_hidden_size": 128,
|
| 16 |
+
"projection_hidden_size": 768
|
| 17 |
+
},
|
| 18 |
+
"hidden_size": 768,
|
| 19 |
+
"initializer_factor": 1.0,
|
| 20 |
+
"logit_scale_init_value": 14.285714285714285,
|
| 21 |
+
"model_type": "clap",
|
| 22 |
+
"num_hidden_layers": 16,
|
| 23 |
+
"projection_dim": 512,
|
| 24 |
+
"projection_hidden_act": "relu",
|
| 25 |
+
"text_config": {
|
| 26 |
+
"classifier_dropout": null,
|
| 27 |
+
"fusion_hidden_size": 768,
|
| 28 |
+
"fusion_num_hidden_layers": 2,
|
| 29 |
+
"initializer_range": 0.02,
|
| 30 |
+
"model_type": "clap_text_model",
|
| 31 |
+
"projection_hidden_size": 768
|
| 32 |
+
},
|
| 33 |
+
"torch_dtype": "float64",
|
| 34 |
+
"transformers_version": "4.32.0.dev0"
|
| 35 |
+
}
|