Upload config.json with huggingface_hub
Browse files- config.json +40 -0
config.json
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_type": "gem",
|
| 3 |
+
"body_model": "soma",
|
| 4 |
+
"denoiser": {
|
| 5 |
+
"latent_dim": 1024,
|
| 6 |
+
"num_layers": 16,
|
| 7 |
+
"num_heads": 8,
|
| 8 |
+
"mlp_ratio": 4,
|
| 9 |
+
"scale_params_dim": 69,
|
| 10 |
+
"scale_comps_dim": 28,
|
| 11 |
+
"fix_scale_comps": true,
|
| 12 |
+
"encode_text": false,
|
| 13 |
+
"encoded_text_dim": 4096,
|
| 14 |
+
"text_encoder_cfg": {
|
| 15 |
+
"mode": "none",
|
| 16 |
+
"cross_attn_type": "mha"
|
| 17 |
+
},
|
| 18 |
+
"text_mask_prob": 0.1,
|
| 19 |
+
"music_mask_prob": 0.1,
|
| 20 |
+
"speech_mask_prob": 0.1,
|
| 21 |
+
"use_text_pos_enc": true
|
| 22 |
+
},
|
| 23 |
+
"endecoder": {
|
| 24 |
+
"stats_name": "MM_V2_SOMA_METROSIM",
|
| 25 |
+
"encode_type": "soma_v2",
|
| 26 |
+
"feat_dim": 585,
|
| 27 |
+
"clip_std": true
|
| 28 |
+
},
|
| 29 |
+
"diffusion": {
|
| 30 |
+
"sampler": "ddim",
|
| 31 |
+
"noise_schedule": "cosine",
|
| 32 |
+
"test_timestep_respacing": "50",
|
| 33 |
+
"ddim_eta": 0.0,
|
| 34 |
+
"guidance_param": 2.5
|
| 35 |
+
},
|
| 36 |
+
"pipeline": {
|
| 37 |
+
"remove_fingers": true,
|
| 38 |
+
"extract_features": false
|
| 39 |
+
}
|
| 40 |
+
}
|