Archive Kurome classifier: AnatomyFlaws-v11_adabelief_fl_naflex_3000.config.json

kurome/classifiers/anatomy-flaws/AnatomyFlaws-v11_adabelief_fl_naflex_3000.config.json

@@ -0,0 +1,70 @@
+{
+  "data_root": "data",
+  "wandb_project": "lumi-classifiers",
+  "resume": null,
+  "base": "AnatomyFlaws",
+  "rev": "v11_adabelief_fl_naflex_3000",
+  "arch": "class",
+  "name": "AnatomyFlaws-v11_adabelief_fl_naflex_3000",
+  "base_vision_model": "google/siglip2-so400m-patch16-naflex",
+  "embed_ver": "siglip2_so400m_patch16_naflex_Naflex_Proc1024",
+  "data_mode": "embeddings",
+  "is_end_to_end": false,
+  "features": 1152,
+  "preload_data": true,
+  "head_features": 1152,
+  "head_hidden_dim": 1280,
+  "pooling_strategy": "none",
+  "head_num_res_blocks": 3,
+  "head_dropout_rate": 0.218,
+  "head_output_mode": "sigmoid",
+  "attn_pool_heads": 16,
+  "attn_pool_dropout": 0.218,
+  "lr": "1e-4",
+  "batch": 32,
+  "loss_function": "focal",
+  "optimizer": "adabelief",
+  "betas": [
+    0.9,
+    0.95
+  ],
+  "eps": "1e-8",
+  "weight_decay": "1e-3",
+  "max_train_epochs": null,
+  "max_train_steps": 10000,
+  "precision": "fp32",
+  "nsave": 1000,
+  "val_split_count": 300,
+  "seed": 218,
+  "num_workers": 0,
+  "save_full_model": false,
+  "log_every_n": 10,
+  "validate_every_n": 100,
+  "focal_loss_gamma": 2.0,
+  "rectify": true,
+  "weight_decouple": true,
+  "gamma": 0.005,
+  "r_sf": 0.0,
+  "wlpow_sf": 2.0,
+  "state_precision": "parameter",
+  "adaptive_clip": 1.0,
+  "adaptive_clip_eps": "1e-3",
+  "scheduler_name": "RexAnnealingWarmRestarts",
+  "scheduler_gamma": 0.95,
+  "scheduler_cycle_multiplier": 1.0,
+  "scheduler_first_cycle_max_steps": 3000,
+  "scheduler_min_lr": "1e-7",
+  "scheduler_warmup_steps": 200,
+  "labels": {
+    "0": "Bad Anatomy",
+    "1": "Good Anatomy"
+  },
+  "weights": [
+    1.0,
+    1.0
+  ],
+  "num_labels": 2,
+  "steps_per_epoch": 75,
+  "num_train_epochs": 134,
+  "num_classes": 2
+}