Add htdemucs_config.json
Browse files- htdemucs_config.json +76 -0
htdemucs_config.json
ADDED
|
@@ -0,0 +1,76 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"klass": "<class 'demucs.htdemucs.HTDemucs'>",
|
| 3 |
+
"args": [],
|
| 4 |
+
"kwargs": {
|
| 5 |
+
"sources": [
|
| 6 |
+
"drums",
|
| 7 |
+
"bass",
|
| 8 |
+
"other",
|
| 9 |
+
"vocals"
|
| 10 |
+
],
|
| 11 |
+
"audio_channels": 2,
|
| 12 |
+
"samplerate": 44100,
|
| 13 |
+
"channels": 48,
|
| 14 |
+
"channels_time": null,
|
| 15 |
+
"growth": 2,
|
| 16 |
+
"nfft": 4096,
|
| 17 |
+
"wiener_iters": 0,
|
| 18 |
+
"end_iters": 0,
|
| 19 |
+
"wiener_residual": false,
|
| 20 |
+
"cac": true,
|
| 21 |
+
"depth": 4,
|
| 22 |
+
"rewrite": true,
|
| 23 |
+
"multi_freqs": [],
|
| 24 |
+
"multi_freqs_depth": 3,
|
| 25 |
+
"freq_emb": 0.2,
|
| 26 |
+
"emb_scale": 10,
|
| 27 |
+
"emb_smooth": true,
|
| 28 |
+
"kernel_size": 8,
|
| 29 |
+
"stride": 4,
|
| 30 |
+
"time_stride": 2,
|
| 31 |
+
"context": 1,
|
| 32 |
+
"context_enc": 0,
|
| 33 |
+
"norm_starts": 4,
|
| 34 |
+
"norm_groups": 4,
|
| 35 |
+
"dconv_mode": 3,
|
| 36 |
+
"dconv_depth": 2,
|
| 37 |
+
"dconv_comp": 8,
|
| 38 |
+
"dconv_init": 0.001,
|
| 39 |
+
"bottom_channels": 512,
|
| 40 |
+
"t_layers": 5,
|
| 41 |
+
"t_hidden_scale": 4.0,
|
| 42 |
+
"t_heads": 8,
|
| 43 |
+
"t_dropout": 0.02,
|
| 44 |
+
"t_layer_scale": true,
|
| 45 |
+
"t_gelu": true,
|
| 46 |
+
"t_emb": "sin",
|
| 47 |
+
"t_max_positions": 10000,
|
| 48 |
+
"t_max_period": 10000.0,
|
| 49 |
+
"t_weight_pos_embed": 1.0,
|
| 50 |
+
"t_cape_mean_normalize": true,
|
| 51 |
+
"t_cape_augment": true,
|
| 52 |
+
"t_cape_glob_loc_scale": [
|
| 53 |
+
5000.0,
|
| 54 |
+
1.0,
|
| 55 |
+
1.4
|
| 56 |
+
],
|
| 57 |
+
"t_sin_random_shift": 0,
|
| 58 |
+
"t_norm_in": true,
|
| 59 |
+
"t_norm_in_group": false,
|
| 60 |
+
"t_group_norm": false,
|
| 61 |
+
"t_norm_first": true,
|
| 62 |
+
"t_norm_out": true,
|
| 63 |
+
"t_weight_decay": 0.0,
|
| 64 |
+
"t_lr": null,
|
| 65 |
+
"t_sparse_self_attn": false,
|
| 66 |
+
"t_sparse_cross_attn": false,
|
| 67 |
+
"t_mask_type": "diag",
|
| 68 |
+
"t_mask_random_seed": 42,
|
| 69 |
+
"t_sparse_attn_window": 400,
|
| 70 |
+
"t_global_window": 100,
|
| 71 |
+
"t_sparsity": 0.95,
|
| 72 |
+
"t_auto_sparsity": false,
|
| 73 |
+
"t_cross_first": false,
|
| 74 |
+
"rescale": 0.1
|
| 75 |
+
}
|
| 76 |
+
}
|