Upload vil_tracker/configs/default.json with huggingface_hub
Browse files
vil_tracker/configs/default.json
ADDED
|
@@ -0,0 +1,53 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model": {
|
| 3 |
+
"dim": 384,
|
| 4 |
+
"depth": 24,
|
| 5 |
+
"patch_size": 16,
|
| 6 |
+
"proj_factor": 2.0,
|
| 7 |
+
"qkv_proj_blocksize": 4,
|
| 8 |
+
"num_heads": 4,
|
| 9 |
+
"conv_kernel": 4,
|
| 10 |
+
"mlp_ratio": 4.0,
|
| 11 |
+
"drop_path_rate": 0.1,
|
| 12 |
+
"tmoe_blocks": 2,
|
| 13 |
+
"num_experts": 4,
|
| 14 |
+
"film_interval": 6,
|
| 15 |
+
"feat_size": 16,
|
| 16 |
+
"use_uncertainty": true
|
| 17 |
+
},
|
| 18 |
+
"training": {
|
| 19 |
+
"phase1": {
|
| 20 |
+
"epochs": 300,
|
| 21 |
+
"lr": 1e-4,
|
| 22 |
+
"batch_size": 32,
|
| 23 |
+
"weight_decay": 0.05,
|
| 24 |
+
"warmup_epochs": 5,
|
| 25 |
+
"grad_clip": 1.0,
|
| 26 |
+
"backbone_lr_scale": 0.1,
|
| 27 |
+
"acl_ramp_epochs": 50
|
| 28 |
+
},
|
| 29 |
+
"phase2": {
|
| 30 |
+
"epochs": 100,
|
| 31 |
+
"lr": 1e-5,
|
| 32 |
+
"batch_size": 32,
|
| 33 |
+
"backbone_lr_scale": 0.01,
|
| 34 |
+
"warmup_epochs": 2
|
| 35 |
+
}
|
| 36 |
+
},
|
| 37 |
+
"data": {
|
| 38 |
+
"template_size": 128,
|
| 39 |
+
"search_size": 256,
|
| 40 |
+
"datasets": ["got10k", "lasot", "trackingnet", "coco"]
|
| 41 |
+
},
|
| 42 |
+
"inference": {
|
| 43 |
+
"search_scale": 4.0,
|
| 44 |
+
"confidence_threshold": 0.3,
|
| 45 |
+
"template_update_threshold": 0.8
|
| 46 |
+
},
|
| 47 |
+
"constraints": {
|
| 48 |
+
"max_params_M": 50,
|
| 49 |
+
"max_latency_ms": 30,
|
| 50 |
+
"max_gflops": 20,
|
| 51 |
+
"max_size_mb": 500
|
| 52 |
+
}
|
| 53 |
+
}
|