Update david_config.json - Run 20251104_140742
Browse files
weights/David-partial_shared-cantor_scale/20251104_140742/david_config.json
ADDED
|
@@ -0,0 +1,50 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"name": "david_clip_vit_bigg14_cantor",
|
| 3 |
+
"uid": "c.david.clip_vit_bigg14_cantor",
|
| 4 |
+
"feature_dim": 1280,
|
| 5 |
+
"num_classes": 1000,
|
| 6 |
+
"scales": [
|
| 7 |
+
384,
|
| 8 |
+
512,
|
| 9 |
+
768,
|
| 10 |
+
1024,
|
| 11 |
+
1280,
|
| 12 |
+
1536,
|
| 13 |
+
1792,
|
| 14 |
+
2048
|
| 15 |
+
],
|
| 16 |
+
"sharing_mode": "partial_shared",
|
| 17 |
+
"fusion_mode": "cantor_scale",
|
| 18 |
+
"use_belly": true,
|
| 19 |
+
"belly_expand": 2.0,
|
| 20 |
+
"projection_temperature": 0.07,
|
| 21 |
+
"shared_feature_dim": 1536,
|
| 22 |
+
"shared_layers": 2,
|
| 23 |
+
"shared_dropout": 0.1,
|
| 24 |
+
"fusion_temperature": 1.0,
|
| 25 |
+
"fusion_dropout": 0.1,
|
| 26 |
+
"tree_depth": 3,
|
| 27 |
+
"num_experts": 3,
|
| 28 |
+
"compression_ratio": 4,
|
| 29 |
+
"expert_dropout": 0.1,
|
| 30 |
+
"attention_dropout": 0.1,
|
| 31 |
+
"geometric_num_heads": 4,
|
| 32 |
+
"geometric_use_cayley": true,
|
| 33 |
+
"geometric_use_angular": true,
|
| 34 |
+
"geometric_scale_dim_aware": true,
|
| 35 |
+
"cantor_num_heads": 8,
|
| 36 |
+
"cantor_depth": 10,
|
| 37 |
+
"cantor_local_window": 5,
|
| 38 |
+
"cantor_use_scale_embeddings": true,
|
| 39 |
+
"progressive_training": false,
|
| 40 |
+
"scale_warmup_epochs": {
|
| 41 |
+
"384": 0,
|
| 42 |
+
"512": 0,
|
| 43 |
+
"768": 0,
|
| 44 |
+
"1024": 0,
|
| 45 |
+
"1280": 0,
|
| 46 |
+
"1536": 0,
|
| 47 |
+
"1792": 0,
|
| 48 |
+
"2048": 0
|
| 49 |
+
}
|
| 50 |
+
}
|