Update david_config.json - Run 20251013_004438
Browse files
weights/David-decoupled-deep_efficiency/20251013_004438/david_config.json
ADDED
|
@@ -0,0 +1,43 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"name": "david_gated_expert_team",
|
| 3 |
+
"uid": "c.david.gated_expert_team",
|
| 4 |
+
"feature_dim": 512,
|
| 5 |
+
"num_classes": 1000,
|
| 6 |
+
"scales": [
|
| 7 |
+
128,
|
| 8 |
+
256,
|
| 9 |
+
384,
|
| 10 |
+
448,
|
| 11 |
+
512,
|
| 12 |
+
576,
|
| 13 |
+
640,
|
| 14 |
+
768,
|
| 15 |
+
896
|
| 16 |
+
],
|
| 17 |
+
"sharing_mode": "decoupled",
|
| 18 |
+
"fusion_mode": "deep_efficiency",
|
| 19 |
+
"use_belly": true,
|
| 20 |
+
"belly_expand": 4,
|
| 21 |
+
"shared_feature_dim": 1024,
|
| 22 |
+
"shared_layers": 4,
|
| 23 |
+
"shared_dropout": 0.1,
|
| 24 |
+
"fusion_temperature": 1.0,
|
| 25 |
+
"fusion_dropout": 0.1,
|
| 26 |
+
"tree_depth": 3,
|
| 27 |
+
"num_experts": 8,
|
| 28 |
+
"compression_ratio": 2,
|
| 29 |
+
"expert_dropout": 0.1,
|
| 30 |
+
"attention_dropout": 0.1,
|
| 31 |
+
"progressive_training": true,
|
| 32 |
+
"scale_warmup_epochs": {
|
| 33 |
+
"128": 0,
|
| 34 |
+
"256": 0,
|
| 35 |
+
"384": 0,
|
| 36 |
+
"448": 0,
|
| 37 |
+
"512": 0,
|
| 38 |
+
"576": 0,
|
| 39 |
+
"640": 0,
|
| 40 |
+
"768": 0,
|
| 41 |
+
"896": 0
|
| 42 |
+
}
|
| 43 |
+
}
|