Upload FlashJudge artifacts from run_20260428_111358

flash_judge_checkpoint.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8ce7ad46d9f629a2862024d84efbf474beb5ecee91bef6e992bf82092d3b545f
+size 1534188467

training_config.json

@@ -0,0 +1,69 @@
+{
+  "model_id": "google/siglip2-base-patch16-naflex",
+  "dataset": "apartments",
+  "apartments_jsonl": "data/apartments/question-image-dataset_labeled_expanded.jsonl",
+  "apartments_images_dir": "data/apartments/images",
+  "fmnist_root": "data",
+  "output_root": "models",
+  "run_name": null,
+  "batch_size": 20,
+  "gradient_accumulation_steps": 1,
+  "stack_size": 1,
+  "epochs": 5,
+  "max_steps": null,
+  "eval_interval": 1,
+  "eval_batches": 2,
+  "checkpoint_interval": 1000,
+  "checkpoint_mode": "max_val_acc",
+  "val_ratio": 0.1,
+  "seed": 42,
+  "apartments_train_group_limit": null,
+  "apartments_train_question_limit": null,
+  "apartments_question_split_mode": "disjoint",
+  "apartments_hard_negative_strategy": "within_group_overlap",
+  "apartments_hard_negative_topk": 0,
+  "apartments_hard_negative_oversample_factor": 0,
+  "num_workers_train": 4,
+  "num_workers_val": 2,
+  "text_max_length": 64,
+  "image_max_num_patches": null,
+  "lr_backbone": 1e-6,
+  "lr_head": 0.0002,
+  "weight_decay": 0.01,
+  "freeze_backbone": false,
+  "freeze_backbone_epochs": 5,
+  "freeze_backbone_steps": null,
+  "backbone_trainable_scope": "full",
+  "attention_heads": 2,
+  "attention_dropout": 0.3,
+  "image_self_attention_layers": 0,
+  "aggregation_mode": "attention",
+  "head_feature_mode": "matching",
+  "late_interaction_topk": 8,
+  "head_hidden_dim": 768,
+  "head_num_layers": 2,
+  "head_dropout": 0.2,
+  "head_activation": "swiglu",
+  "label_smoothing": 0.05,
+  "alignment_loss_weight": 0.05,
+  "alignment_loss_logit_scale": 18.0,
+  "alignment_loss_hard_negative_topk": 1,
+  "compile_model": false,
+  "compile_fallback_to_eager": true,
+  "backbone_torch_dtype": "auto",
+  "attn_implementation": null,
+  "resume_path": null,
+  "min_free_space_gb": 1.0,
+  "use_wandb": true,
+  "wandb_project": "flash-judge",
+  "wandb_entity": null,
+  "wandb_run_name": null,
+  "save_artifacts": true,
+  "push_to_hub": true,
+  "hub_org": "FuncAI",
+  "hub_repo_name": "FlashJudgev5_frozen",
+  "hub_private": false,
+  "hub_token": null,
+  "device": null,
+  "use_amp": true
+}

training_logs.csv

@@ -0,0 +1,6 @@
+epoch,train_loss,train_acc,val_loss,val_acc,max_val_acc
+1,0.5482786992188536,0.7919552711904203,0.44720758710316094,0.7876536956392839,0.7876536956392839
+2,0.4527545233316229,0.8569891210769351,0.49181961134022684,0.7818391476681487,0.7876536956392839
+3,0.3874157649465124,0.8966062064647413,0.5176910986688721,0.7817879174759642,0.7876536956392839
+4,0.3331221164834465,0.9286364169992956,0.5424434945223945,0.7795594336739817,0.7876536956392839
+5,0.29440113875035007,0.9511133286373953,0.581797330123067,0.7810194752499705,0.7876536956392839