| { | |
| "_name_or_path": "/lv0/test_aesthetics/trainlib/projects/aesthetics/experiments/ex04_iterative_classifier_train/distill-n4_00-01_combined_cls_v1b3_siglip2_focal-loss/checkpoint-670", | |
| "architectures": [ | |
| "SiglipForImageClassification" | |
| ], | |
| "finetuning_task": "image-classification", | |
| "id2label": { | |
| "0": "star_0", | |
| "1": "star_1" | |
| }, | |
| "initializer_factor": 1.0, | |
| "label2id": { | |
| "star_0": 0, | |
| "star_1": 1 | |
| }, | |
| "model_type": "siglip", | |
| "problem_type": "single_label_classification", | |
| "text_config": { | |
| "hidden_size": 1024, | |
| "intermediate_size": 4096, | |
| "model_type": "siglip_text_model", | |
| "num_attention_heads": 16, | |
| "num_hidden_layers": 24, | |
| "projection_size": 1024, | |
| "torch_dtype": "float32", | |
| "vocab_size": 256000 | |
| }, | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.49.0", | |
| "vision_config": { | |
| "hidden_size": 1024, | |
| "image_size": 512, | |
| "intermediate_size": 4096, | |
| "model_type": "siglip_vision_model", | |
| "num_attention_heads": 16, | |
| "num_hidden_layers": 24, | |
| "torch_dtype": "float32" | |
| } | |
| } | |