danelcsb/sam2.1_hiera_tiny
11.4k
1{2 "architectures": [3 "Sam2VideoModel"4 ],5 "enable_occlusion_spatial_embedding": true,6 "enable_temporal_pos_encoding_for_object_pointers": true,7 "image_size": 1024,8 "initializer_range": 0.02,9 "mask_decoder_config": {10 "attention_downsample_rate": 2,11 "dynamic_multimask_stability_delta": 0.05,12 "dynamic_multimask_stability_thresh": 0.98,13 "dynamic_multimask_via_stability": true,14 "hidden_act": "gelu",15 "hidden_size": 256,16 "iou_head_depth": 3,17 "iou_head_hidden_dim": 256,18 "mlp_dim": 2048,19 "model_type": "",20 "num_attention_heads": 8,21 "num_hidden_layers": 2,22 "num_multimask_outputs": 323 },24 "mask_downsampler_embed_dim": 256,25 "mask_downsampler_hidden_act": "gelu",26 "mask_downsampler_kernel_size": 3,27 "mask_downsampler_padding": 1,28 "mask_downsampler_stride": 2,29 "mask_downsampler_total_stride": 16,30 "max_object_pointers_in_encoder": 16,31 "memory_attention_downsample_rate": 1,32 "memory_attention_dropout": 0.1,33 "memory_attention_feed_forward_hidden_act": "relu",34 "memory_attention_feed_forward_hidden_size": 2048,35 "memory_attention_hidden_size": 256,36 "memory_attention_num_attention_heads": 1,37 "memory_attention_num_layers": 4,38 "memory_attention_rope_dropout": 0.1,39 "memory_attention_rope_feat_sizes": [40 64,41 6442 ],43 "memory_attention_rope_theta": 10000,44 "memory_encoder_hidden_size": 256,45 "memory_encoder_output_channels": 64,46 "memory_fuser_embed_dim": 256,47 "memory_fuser_hidden_act": "gelu",48 "memory_fuser_intermediate_dim": 1024,49 "memory_fuser_kernel_size": 7,50 "memory_fuser_layer_scale_init_value": 1e-06,51 "memory_fuser_num_layers": 2,52 "memory_fuser_padding": 3,53 "model_type": "sam2_video",54 "multimask_max_pt_num": 1,55 "multimask_min_pt_num": 0,56 "multimask_output_for_tracking": true,57 "multimask_output_in_sam": true,58 "num_maskmem": 7,59 "prompt_encoder_config": {60 "hidden_act": "gelu",61 "hidden_size": 256,62 "image_size": 1024,63 "layer_norm_eps": 1e-06,64 "mask_input_channels": 16,65 "model_type": "",66 "num_point_embeddings": 4,67 "patch_size": 16,68 "scale": 169 },70 "sigmoid_bias_for_mem_enc": -10.0,71 "sigmoid_scale_for_mem_enc": 20.0,72 "torch_dtype": "float32",73 "transformers_version": "4.56.0.dev0",74 "vision_config": {75 "backbone_channel_list": [76 768,77 384,78 192,79 9680 ],81 "backbone_config": {82 "_name_or_path": "",83 "add_cross_attention": false,84 "architectures": null,85 "bad_words_ids": null,86 "begin_suppress_tokens": null,87 "blocks_per_stage": [88 1,89 2,90 7,91 292 ],93 "bos_token_id": null,94 "chunk_size_feed_forward": 0,95 "cross_attention_hidden_size": null,96 "decoder_start_token_id": null,97 "diversity_penalty": 0.0,98 "do_sample": false,99 "early_stopping": false,100 "embed_dim_per_stage": [101 96,102 192,103 384,104 768105 ],106 "encoder_no_repeat_ngram_size": 0,107 "eos_token_id": null,108 "exponential_decay_length_penalty": null,109 "finetuning_task": null,110 "forced_bos_token_id": null,111 "forced_eos_token_id": null,112 "global_attention_blocks": [113 5,114 7,115 9116 ],117 "hidden_act": "gelu",118 "hidden_size": 96,119 "id2label": {120 "0": "LABEL_0",121 "1": "LABEL_1"122 },123 "image_size": [124 1024,125 1024126 ],127 "initializer_range": 0.02,128 "is_decoder": false,129 "is_encoder_decoder": false,130 "label2id": {131 "LABEL_0": 0,132 "LABEL_1": 1133 },134 "layer_norm_eps": 1e-06,135 "length_penalty": 1.0,136 "max_length": 20,137 "min_length": 0,138 "mlp_ratio": 4.0,139 "model_type": "sam2_hiera_det_model",140 "no_repeat_ngram_size": 0,141 "num_attention_heads": 1,142 "num_attention_heads_per_stage": [143 1,144 2,145 4,146 8147 ],148 "num_beam_groups": 1,149 "num_beams": 1,150 "num_channels": 3,151 "num_query_pool_stages": 3,152 "num_return_sequences": 1,153 "output_attentions": false,154 "output_hidden_states": false,155 "output_scores": false,156 "pad_token_id": null,157 "patch_kernel_size": [158 7,159 7160 ],161 "patch_padding": [162 3,163 3164 ],165 "patch_stride": [166 4,167 4168 ],169 "prefix": null,170 "problem_type": null,171 "pruned_heads": {},172 "query_stride": [173 2,174 2175 ],176 "remove_invalid_values": false,177 "repetition_penalty": 1.0,178 "return_dict": true,179 "return_dict_in_generate": false,180 "sep_token_id": null,181 "suppress_tokens": null,182 "task_specific_params": null,183 "temperature": 1.0,184 "tf_legacy_loss": false,185 "tie_encoder_decoder": false,186 "tie_word_embeddings": true,187 "tokenizer_class": null,188 "top_k": 50,189 "top_p": 1.0,190 "torch_dtype": null,191 "torchscript": false,192 "typical_p": 1.0,193 "use_bfloat16": false,194 "window_positional_embedding_background_size": [195 7,196 7197 ],198 "window_size_per_stage": [199 8,200 4,201 14,202 7203 ]204 },205 "backbone_feature_sizes": [206 [207 256,208 256209 ],210 [211 128,212 128213 ],214 [215 64,216 64217 ]218 ],219 "fpn_hidden_size": 256,220 "fpn_kernel_size": 1,221 "fpn_padding": 0,222 "fpn_stride": 1,223 "fpn_top_down_levels": [224 2,225 3226 ],227 "hidden_act": "gelu",228 "initializer_range": 0.02,229 "layer_norm_eps": 1e-06,230 "model_type": "sam2_vision_model",231 "num_feature_levels": 3232 }233}234 