Upload config.json
Browse files- config.json +113 -0
config.json
ADDED
|
@@ -0,0 +1,113 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_type": "GenoLiteHybrid",
|
| 3 |
+
|
| 4 |
+
"vocab_size": 4,
|
| 5 |
+
"sequence_length": 64,
|
| 6 |
+
"num_classes": 3,
|
| 7 |
+
|
| 8 |
+
"d_model": 512,
|
| 9 |
+
|
| 10 |
+
"cnn": {
|
| 11 |
+
"enabled": true,
|
| 12 |
+
"blocks": 7,
|
| 13 |
+
"channels": 960,
|
| 14 |
+
"kernel_size": 3,
|
| 15 |
+
"residual": true,
|
| 16 |
+
"layernorm": true,
|
| 17 |
+
"activation": "gelu"
|
| 18 |
+
},
|
| 19 |
+
|
| 20 |
+
"gru": {
|
| 21 |
+
"enabled": true,
|
| 22 |
+
"hidden_size": 960,
|
| 23 |
+
"layers": 4,
|
| 24 |
+
"bidirectional": false,
|
| 25 |
+
"batch_first": true,
|
| 26 |
+
"projection_to_d_model": true,
|
| 27 |
+
"layernorm": true
|
| 28 |
+
},
|
| 29 |
+
|
| 30 |
+
"transformer": {
|
| 31 |
+
"enabled": true,
|
| 32 |
+
"layers": 6,
|
| 33 |
+
"heads": 8,
|
| 34 |
+
"ffn_dim": 2048,
|
| 35 |
+
"dropout": 0.1,
|
| 36 |
+
"activation": "gelu",
|
| 37 |
+
"batch_first": true,
|
| 38 |
+
"layernorm": true
|
| 39 |
+
},
|
| 40 |
+
|
| 41 |
+
"mamba": {
|
| 42 |
+
"enabled": true,
|
| 43 |
+
"layers": 10,
|
| 44 |
+
"state_dim": 1408,
|
| 45 |
+
"gated": true,
|
| 46 |
+
"residual": true,
|
| 47 |
+
"layernorm": true
|
| 48 |
+
},
|
| 49 |
+
|
| 50 |
+
"fusion": {
|
| 51 |
+
"input_dim": 2048,
|
| 52 |
+
"output_dim": 512,
|
| 53 |
+
"activation": "gelu",
|
| 54 |
+
"dropout": 0.1,
|
| 55 |
+
"layernorm": true
|
| 56 |
+
},
|
| 57 |
+
|
| 58 |
+
"classifier": {
|
| 59 |
+
"hidden_dim": 512,
|
| 60 |
+
"dropout": 0.1,
|
| 61 |
+
"activation": "gelu",
|
| 62 |
+
"num_classes": 3
|
| 63 |
+
},
|
| 64 |
+
|
| 65 |
+
"pooling": {
|
| 66 |
+
"type": "mean"
|
| 67 |
+
},
|
| 68 |
+
|
| 69 |
+
"training": {
|
| 70 |
+
"epochs": 3,
|
| 71 |
+
"batch_size": 3,
|
| 72 |
+
"learning_rate": 0.0001,
|
| 73 |
+
"optimizer": "AdamW",
|
| 74 |
+
"weight_decay": 0.01,
|
| 75 |
+
"gradient_clipping": 1.0,
|
| 76 |
+
"shuffle": true
|
| 77 |
+
},
|
| 78 |
+
|
| 79 |
+
"dataset": {
|
| 80 |
+
"type": "synthetic",
|
| 81 |
+
"samples_total": 9000,
|
| 82 |
+
"samples_per_class": 3000,
|
| 83 |
+
|
| 84 |
+
"classes": [
|
| 85 |
+
"OK",
|
| 86 |
+
"MHAP",
|
| 87 |
+
"PROBLEM"
|
| 88 |
+
],
|
| 89 |
+
|
| 90 |
+
"difficulty_levels": [
|
| 91 |
+
"easy",
|
| 92 |
+
"medium",
|
| 93 |
+
"hard"
|
| 94 |
+
],
|
| 95 |
+
|
| 96 |
+
"features": [
|
| 97 |
+
"controlled_entropy",
|
| 98 |
+
"motif_repetition",
|
| 99 |
+
"hidden_illegal_pairs",
|
| 100 |
+
"partial_shuffle",
|
| 101 |
+
"duplicate_prevention",
|
| 102 |
+
"class_overlap"
|
| 103 |
+
]
|
| 104 |
+
},
|
| 105 |
+
|
| 106 |
+
"hardware": {
|
| 107 |
+
"device": "cpu",
|
| 108 |
+
"ram_gb": 8,
|
| 109 |
+
"cpu": "Intel i7-4700MQ"
|
| 110 |
+
},
|
| 111 |
+
|
| 112 |
+
"estimated_parameters": "88M+"
|
| 113 |
+
}
|