Create model_config.json
Browse files- model_config.json +179 -0
model_config.json
ADDED
|
@@ -0,0 +1,179 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_type": "diffusion_cond",
|
| 3 |
+
"sample_size": 2097152,
|
| 4 |
+
"sample_rate": 44100,
|
| 5 |
+
"audio_channels": 2,
|
| 6 |
+
"model": {
|
| 7 |
+
"pretransform": {
|
| 8 |
+
"type": "autoencoder",
|
| 9 |
+
"iterate_batch": true,
|
| 10 |
+
"config": {
|
| 11 |
+
"encoder": {
|
| 12 |
+
"type": "oobleck",
|
| 13 |
+
"requires_grad": false,
|
| 14 |
+
"config": {
|
| 15 |
+
"in_channels": 2,
|
| 16 |
+
"channels": 128,
|
| 17 |
+
"c_mults": [
|
| 18 |
+
1,
|
| 19 |
+
2,
|
| 20 |
+
4,
|
| 21 |
+
8,
|
| 22 |
+
16
|
| 23 |
+
],
|
| 24 |
+
"strides": [
|
| 25 |
+
2,
|
| 26 |
+
4,
|
| 27 |
+
4,
|
| 28 |
+
8,
|
| 29 |
+
8
|
| 30 |
+
],
|
| 31 |
+
"latent_dim": 128,
|
| 32 |
+
"use_snake": true
|
| 33 |
+
}
|
| 34 |
+
},
|
| 35 |
+
"decoder": {
|
| 36 |
+
"type": "oobleck",
|
| 37 |
+
"config": {
|
| 38 |
+
"out_channels": 2,
|
| 39 |
+
"channels": 128,
|
| 40 |
+
"c_mults": [
|
| 41 |
+
1,
|
| 42 |
+
2,
|
| 43 |
+
4,
|
| 44 |
+
8,
|
| 45 |
+
16
|
| 46 |
+
],
|
| 47 |
+
"strides": [
|
| 48 |
+
2,
|
| 49 |
+
4,
|
| 50 |
+
4,
|
| 51 |
+
8,
|
| 52 |
+
8
|
| 53 |
+
],
|
| 54 |
+
"latent_dim": 64,
|
| 55 |
+
"use_snake": true,
|
| 56 |
+
"final_tanh": false
|
| 57 |
+
}
|
| 58 |
+
},
|
| 59 |
+
"bottleneck": {
|
| 60 |
+
"type": "vae"
|
| 61 |
+
},
|
| 62 |
+
"latent_dim": 64,
|
| 63 |
+
"downsampling_ratio": 2048,
|
| 64 |
+
"io_channels": 2
|
| 65 |
+
}
|
| 66 |
+
},
|
| 67 |
+
"conditioning": {
|
| 68 |
+
"configs": [
|
| 69 |
+
{
|
| 70 |
+
"id": "prompt",
|
| 71 |
+
"type": "t5",
|
| 72 |
+
"config": {
|
| 73 |
+
"t5_model_name": "t5-base",
|
| 74 |
+
"max_length": 128
|
| 75 |
+
}
|
| 76 |
+
},
|
| 77 |
+
{
|
| 78 |
+
"id": "seconds_start",
|
| 79 |
+
"type": "number",
|
| 80 |
+
"config": {
|
| 81 |
+
"min_val": 0,
|
| 82 |
+
"max_val": 512
|
| 83 |
+
}
|
| 84 |
+
},
|
| 85 |
+
{
|
| 86 |
+
"id": "seconds_total",
|
| 87 |
+
"type": "number",
|
| 88 |
+
"config": {
|
| 89 |
+
"min_val": 0,
|
| 90 |
+
"max_val": 512
|
| 91 |
+
}
|
| 92 |
+
}
|
| 93 |
+
],
|
| 94 |
+
"cond_dim": 768
|
| 95 |
+
},
|
| 96 |
+
"diffusion": {
|
| 97 |
+
"cross_attention_cond_ids": [
|
| 98 |
+
"prompt",
|
| 99 |
+
"seconds_start",
|
| 100 |
+
"seconds_total"
|
| 101 |
+
],
|
| 102 |
+
"global_cond_ids": [
|
| 103 |
+
"seconds_start",
|
| 104 |
+
"seconds_total"
|
| 105 |
+
],
|
| 106 |
+
"type": "dit",
|
| 107 |
+
"config": {
|
| 108 |
+
"io_channels": 64,
|
| 109 |
+
"embed_dim": 1536,
|
| 110 |
+
"depth": 24,
|
| 111 |
+
"num_heads": 24,
|
| 112 |
+
"cond_token_dim": 768,
|
| 113 |
+
"global_cond_dim": 1536,
|
| 114 |
+
"project_cond_tokens": false,
|
| 115 |
+
"transformer_type": "continuous_transformer"
|
| 116 |
+
}
|
| 117 |
+
},
|
| 118 |
+
"io_channels": 64
|
| 119 |
+
},
|
| 120 |
+
"training": {
|
| 121 |
+
"use_ema": true,
|
| 122 |
+
"log_loss_info": false,
|
| 123 |
+
"optimizer_configs": {
|
| 124 |
+
"diffusion": {
|
| 125 |
+
"optimizer": {
|
| 126 |
+
"type": "AdamW",
|
| 127 |
+
"config": {
|
| 128 |
+
"lr": 5e-5,
|
| 129 |
+
"betas": [
|
| 130 |
+
0.9,
|
| 131 |
+
0.999
|
| 132 |
+
],
|
| 133 |
+
"weight_decay": 1e-3
|
| 134 |
+
}
|
| 135 |
+
},
|
| 136 |
+
"scheduler": {
|
| 137 |
+
"type": "InverseLR",
|
| 138 |
+
"config": {
|
| 139 |
+
"inv_gamma": 1000000,
|
| 140 |
+
"power": 0.5,
|
| 141 |
+
"warmup": 0.99
|
| 142 |
+
}
|
| 143 |
+
}
|
| 144 |
+
}
|
| 145 |
+
},
|
| 146 |
+
"demo": {
|
| 147 |
+
"demo_every": 687,
|
| 148 |
+
"demo_steps": 250,
|
| 149 |
+
"num_demos": 4,
|
| 150 |
+
"demo_cond": [
|
| 151 |
+
{
|
| 152 |
+
"prompt": "Format: Solo | Genre: Trap | Sub-Genre: Cinematic Piano / Ambient | Instruments: Acoustic Piano | Moods: Sad, Melancholic, Reflective, Somber, Introspective | Styles: Expressive, Emotional, Sparse | Tempo: Slow | BPM: 70 | Key: Am",
|
| 153 |
+
"seconds_start": 0,
|
| 154 |
+
"seconds_total": 15
|
| 155 |
+
},
|
| 156 |
+
{
|
| 157 |
+
"prompt": "Format: Solo | Genre: Trap | Sub-Genre: Melodic Trap / Future Pop | Instruments: Synth Lead, Synth Plucks, Atmospheric Pads | Moods: Bright, Uplifting, Carefree, Energetic | Styles: Wavy, Catchy, Modern, Smooth | Tempo: Fast | BPM: 155 | Key: G",
|
| 158 |
+
"seconds_start": 0,
|
| 159 |
+
"seconds_total": 30
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"prompt": "Format: Solo | Genre: Trap | Sub-Genre: Hard Trap / Cinematic Trap | Instruments: Synth Pads (dark & evolving), Sub Bass, FX Risers, Distorted Synth Stabs | Moods: Epic, Deep, Intense, Dark, Anticipatory, Powerful | Styles: Atmospheric, Building, Menacing, Cinematic | Tempo: Medium-Fast | BPM: 150 | Key: Fm",
|
| 163 |
+
"seconds_start": 0,
|
| 164 |
+
"seconds_total": 12
|
| 165 |
+
},
|
| 166 |
+
{
|
| 167 |
+
"prompt": "Format: Solo | Genre: Trap | Sub-Genre: Trap | Instruments: 808 Bass, Programmed Drums, Hi-Hats, Snare | Moods: Dark, Menacing, Energetic, Driving | Styles: Hard-hitting, Rhythmic, Urban, Gritty | Tempo: Mid-Fast, Driving | BPM: 130 | Key: Cm",
|
| 168 |
+
"seconds_start": 0,
|
| 169 |
+
"seconds_total": 18
|
| 170 |
+
}
|
| 171 |
+
],
|
| 172 |
+
"demo_cfg_scales": [
|
| 173 |
+
7,
|
| 174 |
+
9,
|
| 175 |
+
10
|
| 176 |
+
]
|
| 177 |
+
}
|
| 178 |
+
}
|
| 179 |
+
}
|