tiny-random/sam3

1{2  "architectures": [3    "Sam3Model"4  ],5  "detr_decoder_config": {6    "box_rpb_mode": "log",7    "dropout": 0.1,8    "hidden_act": "relu",9    "hidden_dropout": 0.0,10    "hidden_size": 16,11    "initializer_range": 0.02,12    "intermediate_size": 32,13    "layer_norm_eps": 1e-06,14    "model_type": "sam3_detr_decoder",15    "num_attention_heads": 2,16    "num_layers": 6,17    "num_queries": 200,18    "use_presence_token": true19  },20  "detr_encoder_config": {21    "dropout": 0.1,22    "hidden_act": "relu",23    "hidden_dropout": 0.0,24    "hidden_size": 16,25    "initializer_range": 0.02,26    "intermediate_size": 32,27    "layer_norm_eps": 1e-06,28    "model_type": "sam3_detr_encoder",29    "num_attention_heads": 2,30    "num_layers": 631  },32  "dtype": "float32",33  "geometry_encoder_config": {34    "dropout": 0.1,35    "hidden_act": "relu",36    "hidden_dropout": 0.0,37    "hidden_size": 16,38    "initializer_range": 0.02,39    "intermediate_size": 32,40    "layer_norm_eps": 1e-06,41    "model_type": "sam3_geometry_encoder",42    "num_attention_heads": 2,43    "num_layers": 3,44    "roi_size": 745  },46  "initializer_range": 0.02,47  "mask_decoder_config": {48    "dropout": 0.0,49    "hidden_size": 16,50    "initializer_range": 0.02,51    "intermediate_size": 32,52    "layer_norm_eps": 1e-06,53    "model_type": "sam3_mask_decoder",54    "num_attention_heads": 2,55    "num_upsampling_stages": 356  },57  "model_type": "sam3",58  "text_config": {59    "attention_dropout": 0.0,60    "hidden_act": "gelu",61    "hidden_size": 16,62    "initializer_factor": 1.0,63    "initializer_range": 0.02,64    "intermediate_size": 32,65    "layer_norm_eps": 1e-05,66    "max_position_embeddings": 32,67    "model_type": "clip_text_model",68    "num_attention_heads": 2,69    "num_hidden_layers": 2,70    "projection_dim": 16,71    "vocab_size": 4940872  },73  "transformers_version": "5.0.0.dev0",74  "vision_config": {75    "backbone_config": {76      "_name_or_path": "",77      "add_cross_attention": false,78      "architectures": null,79      "attention_dropout": 0.0,80      "bos_token_id": null,81      "chunk_size_feed_forward": 0,82      "cross_attention_hidden_size": null,83      "decoder_start_token_id": null,84      "dtype": null,85      "eos_token_id": null,86      "finetuning_task": null,87      "fpn_hidden_size": 16,88      "global_attn_indexes": [89        1,90        3,91        5,92        793      ],94      "hidden_act": "gelu",95      "hidden_dropout": 0.0,96      "hidden_size": 16,97      "id2label": {98        "0": "LABEL_0",99        "1": "LABEL_1"100      },101      "image_size": 1008,102      "initializer_range": 0.02,103      "intermediate_size": 32,104      "is_decoder": false,105      "is_encoder_decoder": false,106      "label2id": {107        "LABEL_0": 0,108        "LABEL_1": 1109      },110      "layer_norm_eps": 1e-06,111      "layer_scale_init_value": null,112      "model_type": "sam3_vit_model",113      "num_attention_heads": 2,114      "num_channels": 3,115      "num_hidden_layers": 8,116      "output_attentions": false,117      "output_hidden_states": false,118      "pad_token_id": null,119      "patch_size": 14,120      "prefix": null,121      "pretrain_image_size": 336,122      "problem_type": null,123      "qkv_bias": true,124      "return_dict": true,125      "rope_theta": 10000.0,126      "sep_token_id": null,127      "task_specific_params": null,128      "tie_encoder_decoder": false,129      "tie_word_embeddings": true,130      "tokenizer_class": null,131      "window_size": 24132    },133    "backbone_feature_sizes": [134      [135        288,136        288137      ],138      [139        144,140        144141      ],142      [143        72,144        72145      ]146    ],147    "fpn_hidden_size": 16,148    "fpn_kernel_size": 2,149    "fpn_stride": 2,150    "hidden_act": "gelu",151    "initializer_range": 0.02,152    "layer_norm_eps": 1e-06,153    "model_type": "sam3_vision_model",154    "num_feature_levels": 3,155    "scale_factors": [156      4.0,157      2.0,158      1.0,159      0.5160    ]161  }162}163