| { |
| "model": { |
| "clip_model": "openai/clip-vit-large-patch14", |
| "clip_hidden": 768, |
| "clip_layers": 12, |
| "clip_max_tokens": 77, |
| "freeze_clip": true, |
| "n_memory_tokens": 8, |
| "bank_size": 64, |
| "anchor_dim": 768, |
| "n_bank_heads": 8, |
| "bank_cross_layers": 2, |
| "gate_type": "gru", |
| "extract_layers": [ |
| 1, |
| 3, |
| 5, |
| 7, |
| 9, |
| 11 |
| ], |
| "layer_fusion": "learned", |
| "max_content_tokens": 18, |
| "segment_overlap": 4, |
| "max_segments": 32, |
| "cv_target": 0.2, |
| "sequence_output": true, |
| "sequence_len": 77, |
| "sequence_recon_layers": 2, |
| "sequence_recon_heads": 8, |
| "collect_content_tokens": true, |
| "max_content_positions": 256, |
| "teacher_model": "answerdotai/ModernBERT-large", |
| "teacher_hidden": 1024, |
| "return_dict": true, |
| "output_hidden_states": false, |
| "dtype": null, |
| "chunk_size_feed_forward": 0, |
| "is_encoder_decoder": false, |
| "architectures": null, |
| "id2label": { |
| "0": "LABEL_0", |
| "1": "LABEL_1" |
| }, |
| "label2id": { |
| "LABEL_0": 0, |
| "LABEL_1": 1 |
| }, |
| "problem_type": null, |
| "_name_or_path": "", |
| "transformers_version": "5.0.0", |
| "model_type": "memory_clip_seq", |
| "output_attentions": false |
| }, |
| "training": { |
| "max_train_samples": 50000, |
| "max_val_samples": 2000, |
| "min_caption_length": 100, |
| "phase1_epochs": 5, |
| "phase1_lr_seq": 0.002, |
| "phase1_lr_proj": 0.001, |
| "phase2_epochs": 5, |
| "phase2_lr_bank": 0.0005, |
| "phase2_lr_output": 0.0002, |
| "phase2_lr_proj": 0.0005, |
| "phase2_lr_seq": 0.001, |
| "batch_size": 64, |
| "min_lr": 1e-06, |
| "weight_decay": 0.01, |
| "grad_clip": 1.0, |
| "warmup_steps": 200, |
| "modern_weight": 1.0, |
| "procrustes_weight": 0.3, |
| "cv_weight": 0.05, |
| "temperature": 0.07, |
| "sequence_weight": 1.0, |
| "sequence_cosine_weight": 0.5, |
| "modern_max_len": 4096, |
| "procrustes_n_samples": 300, |
| "v1_checkpoint": "", |
| "v1_repo_id": "AbstractPhil/geolip-clip-vit-large-patch14-ctx576", |
| "v1_filename": "model.safetensors", |
| "checkpoint_dir": "/home/claude/memory_clip_seq_checkpoints", |
| "tensorboard_dir": "/home/claude/memory_clip_seq_tb", |
| "metrics_file": "/home/claude/memory_clip_seq_checkpoints/metrics.json", |
| "log_every": 20, |
| "eval_every": 200 |
| } |
| } |
