model.weights.json

{"metadata": {"total_size": 2528505776.25}, "weight_map": {"/layers/conv2d/vars": ["model_00000.weights.h5"], "/layers/vi_t_class_dist_token/vars": "model_00000.weights.h5", "/layers/vi_t_add_position_embs/vars": ["model_00000.weights.h5"], "/layers/layer_normalization/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention/proj/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention/qkv/vars": ["model_00000.weights.h5"], "/layers/layer_scale/vars": "model_00000.weights.h5", "/layers/layer_normalization_1/vars": ["model_00000.weights.h5"], "/layers/dense/vars": ["model_00000.weights.h5"], "/layers/dense_1/vars": ["model_00000.weights.h5"], "/layers/layer_scale_1/vars": "model_00000.weights.h5", "/layers/layer_normalization_2/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_1/proj/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_1/qkv/vars": ["model_00000.weights.h5"], "/layers/layer_scale_2/vars": "model_00000.weights.h5", "/layers/layer_normalization_3/vars": ["model_00000.weights.h5"], "/layers/dense_2/vars": ["model_00000.weights.h5"], "/layers/dense_3/vars": ["model_00000.weights.h5"], "/layers/layer_scale_3/vars": "model_00000.weights.h5", "/layers/layer_normalization_4/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_2/proj/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_2/qkv/vars": ["model_00000.weights.h5"], "/layers/layer_scale_4/vars": "model_00000.weights.h5", "/layers/layer_normalization_5/vars": ["model_00000.weights.h5"], "/layers/dense_4/vars": ["model_00000.weights.h5"], "/layers/dense_5/vars": ["model_00000.weights.h5"], "/layers/layer_scale_5/vars": "model_00000.weights.h5", "/layers/layer_normalization_6/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_3/proj/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_3/qkv/vars": ["model_00000.weights.h5"], "/layers/layer_scale_6/vars": "model_00000.weights.h5", "/layers/layer_normalization_7/vars": ["model_00000.weights.h5"], "/layers/dense_6/vars": ["model_00000.weights.h5"], "/layers/dense_7/vars": ["model_00000.weights.h5"], "/layers/layer_scale_7/vars": "model_00000.weights.h5", "/layers/layer_normalization_8/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_4/proj/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_4/qkv/vars": ["model_00000.weights.h5"], "/layers/layer_scale_8/vars": "model_00000.weights.h5", "/layers/layer_normalization_9/vars": ["model_00000.weights.h5"], "/layers/dense_8/vars": ["model_00000.weights.h5"], "/layers/dense_9/vars": ["model_00000.weights.h5"], "/layers/layer_scale_9/vars": "model_00000.weights.h5", "/layers/layer_normalization_10/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_5/proj/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_5/qkv/vars": ["model_00000.weights.h5"], "/layers/layer_scale_10/vars": "model_00000.weights.h5", "/layers/layer_normalization_11/vars": ["model_00000.weights.h5"], "/layers/dense_10/vars": ["model_00000.weights.h5"], "/layers/dense_11/vars": ["model_00000.weights.h5"], "/layers/layer_scale_11/vars": "model_00000.weights.h5", "/layers/layer_normalization_12/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_6/proj/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_6/qkv/vars": ["model_00000.weights.h5"], "/layers/layer_scale_12/vars": "model_00000.weights.h5", "/layers/layer_normalization_13/vars": ["model_00000.weights.h5"], "/layers/dense_12/vars": ["model_00000.weights.h5"], "/layers/dense_13/vars": ["model_00000.weights.h5"], "/layers/layer_scale_13/vars": "model_00000.weights.h5", "/layers/layer_normalization_14/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_7/proj/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_7/qkv/vars": ["model_00000.weights.h5"], "/layers/layer_scale_14/vars": "model_00000.weights.h5", "/layers/layer_normalization_15/vars": ["model_00000.weights.h5"], "/layers/dense_14/vars": ["model_00000.weights.h5"], "/layers/dense_15/vars": ["model_00000.weights.h5"], "/layers/layer_scale_15/vars": "model_00000.weights.h5", "/layers/layer_normalization_16/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_8/proj/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_8/qkv/vars": ["model_00000.weights.h5"], "/layers/layer_scale_16/vars": "model_00000.weights.h5", "/layers/layer_normalization_17/vars": ["model_00000.weights.h5"], "/layers/dense_16/vars": ["model_00000.weights.h5"], "/layers/dense_17/vars": ["model_00000.weights.h5"], "/layers/layer_scale_17/vars": "model_00000.weights.h5", "/layers/layer_normalization_18/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_9/proj/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_9/qkv/vars": ["model_00000.weights.h5"], "/layers/layer_scale_18/vars": "model_00000.weights.h5", "/layers/layer_normalization_19/vars": ["model_00000.weights.h5"], "/layers/dense_18/vars": ["model_00000.weights.h5"], "/layers/dense_19/vars": ["model_00000.weights.h5"], "/layers/layer_scale_19/vars": "model_00000.weights.h5", "/layers/layer_normalization_20/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_10/proj/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_10/qkv/vars": ["model_00000.weights.h5"], "/layers/layer_scale_20/vars": "model_00000.weights.h5", "/layers/layer_normalization_21/vars": ["model_00000.weights.h5"], "/layers/dense_20/vars": ["model_00000.weights.h5"], "/layers/dense_21/vars": ["model_00000.weights.h5"], "/layers/layer_scale_21/vars": "model_00000.weights.h5", "/layers/layer_normalization_22/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_11/proj/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_11/qkv/vars": ["model_00000.weights.h5"], "/layers/layer_scale_22/vars": "model_00000.weights.h5", "/layers/layer_normalization_23/vars": ["model_00000.weights.h5"], "/layers/dense_22/vars": ["model_00000.weights.h5"], "/layers/dense_23/vars": ["model_00000.weights.h5"], "/layers/layer_scale_23/vars": "model_00000.weights.h5", "/layers/layer_normalization_24/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_12/proj/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_12/qkv/vars": ["model_00000.weights.h5"], "/layers/layer_scale_24/vars": "model_00000.weights.h5", "/layers/layer_normalization_25/vars": ["model_00000.weights.h5"], "/layers/dense_24/vars": ["model_00000.weights.h5"], "/layers/dense_25/vars": ["model_00000.weights.h5"], "/layers/layer_scale_25/vars": "model_00000.weights.h5", "/layers/layer_normalization_26/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_13/proj/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_13/qkv/vars": ["model_00000.weights.h5"], "/layers/layer_scale_26/vars": "model_00000.weights.h5", "/layers/layer_normalization_27/vars": ["model_00000.weights.h5"], "/layers/dense_26/vars": ["model_00000.weights.h5"], "/layers/dense_27/vars": ["model_00000.weights.h5"], "/layers/layer_scale_27/vars": "model_00000.weights.h5", "/layers/layer_normalization_28/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_14/proj/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_14/qkv/vars": ["model_00000.weights.h5"], "/layers/layer_scale_28/vars": "model_00000.weights.h5", "/layers/layer_normalization_29/vars": ["model_00000.weights.h5"], "/layers/dense_28/vars": ["model_00000.weights.h5"], "/layers/dense_29/vars": ["model_00000.weights.h5"], "/layers/layer_scale_29/vars": "model_00000.weights.h5", "/layers/layer_normalization_30/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_15/proj/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_15/qkv/vars": ["model_00000.weights.h5"], "/layers/layer_scale_30/vars": "model_00000.weights.h5", "/layers/layer_normalization_31/vars": ["model_00000.weights.h5"], "/layers/dense_30/vars": ["model_00000.weights.h5"], "/layers/dense_31/vars": ["model_00000.weights.h5"], "/layers/layer_scale_31/vars": "model_00000.weights.h5", "/layers/layer_normalization_32/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_16/proj/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_16/qkv/vars": ["model_00000.weights.h5"], "/layers/layer_scale_32/vars": "model_00000.weights.h5", "/layers/layer_normalization_33/vars": ["model_00000.weights.h5"], "/layers/dense_32/vars": ["model_00000.weights.h5"], "/layers/dense_33/vars": ["model_00000.weights.h5"], "/layers/layer_scale_33/vars": "model_00000.weights.h5", "/layers/layer_normalization_34/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_17/proj/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_17/qkv/vars": ["model_00000.weights.h5"], "/layers/layer_scale_34/vars": "model_00000.weights.h5", "/layers/layer_normalization_35/vars": ["model_00000.weights.h5"], "/layers/dense_34/vars": ["model_00000.weights.h5"], "/layers/dense_35/vars": ["model_00000.weights.h5"], "/layers/layer_scale_35/vars": "model_00000.weights.h5", "/layers/layer_normalization_36/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_18/proj/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_18/qkv/vars": ["model_00000.weights.h5"], "/layers/layer_scale_36/vars": "model_00000.weights.h5", "/layers/layer_normalization_37/vars": ["model_00000.weights.h5"], "/layers/dense_36/vars": ["model_00000.weights.h5"], "/layers/dense_37/vars": ["model_00000.weights.h5"], "/layers/layer_scale_37/vars": "model_00000.weights.h5", "/layers/layer_normalization_38/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_19/proj/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_19/qkv/vars": ["model_00000.weights.h5"], "/layers/layer_scale_38/vars": "model_00000.weights.h5", "/layers/layer_normalization_39/vars": ["model_00000.weights.h5"], "/layers/dense_38/vars": ["model_00000.weights.h5"], "/layers/dense_39/vars": ["model_00000.weights.h5"], "/layers/layer_scale_39/vars": "model_00000.weights.h5", "/layers/layer_normalization_40/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_20/proj/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_20/qkv/vars": ["model_00000.weights.h5"], "/layers/layer_scale_40/vars": "model_00000.weights.h5", "/layers/layer_normalization_41/vars": ["model_00000.weights.h5"], "/layers/dense_40/vars": ["model_00000.weights.h5"], "/layers/dense_41/vars": ["model_00000.weights.h5"], "/layers/layer_scale_41/vars": "model_00000.weights.h5", "/layers/layer_normalization_42/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_21/proj/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_21/qkv/vars": ["model_00000.weights.h5"], "/layers/layer_scale_42/vars": "model_00000.weights.h5", "/layers/layer_normalization_43/vars": ["model_00000.weights.h5"], "/layers/dense_42/vars": ["model_00000.weights.h5"], "/layers/dense_43/vars": ["model_00000.weights.h5"], "/layers/layer_scale_43/vars": "model_00000.weights.h5", "/layers/layer_normalization_44/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_22/proj/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_22/qkv/vars": ["model_00000.weights.h5"], "/layers/layer_scale_44/vars": "model_00000.weights.h5", "/layers/layer_normalization_45/vars": ["model_00000.weights.h5"], "/layers/dense_44/vars": ["model_00000.weights.h5"], "/layers/dense_45/vars": ["model_00000.weights.h5"], "/layers/layer_scale_45/vars": "model_00000.weights.h5", "/layers/layer_normalization_46/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_23/proj/vars": ["model_00000.weights.h5"], "/layers/vi_t_multi_head_self_attention_23/qkv/vars": ["model_00001.weights.h5"], "/layers/layer_scale_46/vars": "model_00001.weights.h5", "/layers/layer_normalization_47/vars": ["model_00001.weights.h5"], "/layers/dense_46/vars": ["model_00001.weights.h5"], "/layers/dense_47/vars": ["model_00001.weights.h5"], "/layers/layer_scale_47/vars": "model_00001.weights.h5", "/layers/layer_normalization_48/vars": ["model_00001.weights.h5"], "/layers/vi_t_multi_head_self_attention_24/proj/vars": ["model_00001.weights.h5"], "/layers/vi_t_multi_head_self_attention_24/qkv/vars": ["model_00001.weights.h5"], "/layers/layer_scale_48/vars": "model_00001.weights.h5", "/layers/layer_normalization_49/vars": ["model_00001.weights.h5"], "/layers/dense_48/vars": ["model_00001.weights.h5"], "/layers/dense_49/vars": ["model_00001.weights.h5"], "/layers/layer_scale_49/vars": "model_00001.weights.h5", "/layers/layer_normalization_50/vars": ["model_00001.weights.h5"], "/layers/vi_t_multi_head_self_attention_25/proj/vars": ["model_00001.weights.h5"], "/layers/vi_t_multi_head_self_attention_25/qkv/vars": ["model_00001.weights.h5"], "/layers/layer_scale_50/vars": "model_00001.weights.h5", "/layers/layer_normalization_51/vars": ["model_00001.weights.h5"], "/layers/dense_50/vars": ["model_00001.weights.h5"], "/layers/dense_51/vars": ["model_00001.weights.h5"], "/layers/layer_scale_51/vars": "model_00001.weights.h5", "/layers/layer_normalization_52/vars": ["model_00001.weights.h5"], "/layers/vi_t_multi_head_self_attention_26/proj/vars": ["model_00001.weights.h5"], "/layers/vi_t_multi_head_self_attention_26/qkv/vars": ["model_00001.weights.h5"], "/layers/layer_scale_52/vars": "model_00001.weights.h5", "/layers/layer_normalization_53/vars": ["model_00001.weights.h5"], "/layers/dense_52/vars": ["model_00001.weights.h5"], "/layers/dense_53/vars": ["model_00001.weights.h5"], "/layers/layer_scale_53/vars": "model_00001.weights.h5", "/layers/layer_normalization_54/vars": ["model_00001.weights.h5"], "/layers/vi_t_multi_head_self_attention_27/proj/vars": ["model_00001.weights.h5"], "/layers/vi_t_multi_head_self_attention_27/qkv/vars": ["model_00001.weights.h5"], "/layers/layer_scale_54/vars": "model_00001.weights.h5", "/layers/layer_normalization_55/vars": ["model_00001.weights.h5"], "/layers/dense_54/vars": ["model_00001.weights.h5"], "/layers/dense_55/vars": ["model_00001.weights.h5"], "/layers/layer_scale_55/vars": "model_00001.weights.h5", "/layers/layer_normalization_56/vars": ["model_00001.weights.h5"], "/layers/vi_t_multi_head_self_attention_28/proj/vars": ["model_00001.weights.h5"], "/layers/vi_t_multi_head_self_attention_28/qkv/vars": ["model_00001.weights.h5"], "/layers/layer_scale_56/vars": "model_00001.weights.h5", "/layers/layer_normalization_57/vars": ["model_00001.weights.h5"], "/layers/dense_56/vars": ["model_00001.weights.h5"], "/layers/dense_57/vars": ["model_00001.weights.h5"], "/layers/layer_scale_57/vars": "model_00001.weights.h5", "/layers/layer_normalization_58/vars": ["model_00001.weights.h5"], "/layers/vi_t_multi_head_self_attention_29/proj/vars": ["model_00001.weights.h5"], "/layers/vi_t_multi_head_self_attention_29/qkv/vars": ["model_00001.weights.h5"], "/layers/layer_scale_58/vars": "model_00001.weights.h5", "/layers/layer_normalization_59/vars": ["model_00001.weights.h5"], "/layers/dense_58/vars": ["model_00001.weights.h5"], "/layers/dense_59/vars": ["model_00001.weights.h5"], "/layers/layer_scale_59/vars": "model_00001.weights.h5", "/layers/layer_normalization_60/vars": ["model_00001.weights.h5"], "/layers/vi_t_multi_head_self_attention_30/proj/vars": ["model_00001.weights.h5"], "/layers/vi_t_multi_head_self_attention_30/qkv/vars": ["model_00001.weights.h5"], "/layers/layer_scale_60/vars": "model_00001.weights.h5", "/layers/layer_normalization_61/vars": ["model_00001.weights.h5"], "/layers/dense_60/vars": ["model_00001.weights.h5"], "/layers/dense_61/vars": ["model_00001.weights.h5"], "/layers/layer_scale_61/vars": "model_00001.weights.h5", "/layers/layer_normalization_62/vars": ["model_00001.weights.h5"], "/layers/vi_t_multi_head_self_attention_31/proj/vars": ["model_00001.weights.h5"], "/layers/vi_t_multi_head_self_attention_31/qkv/vars": ["model_00001.weights.h5"], "/layers/layer_scale_62/vars": "model_00001.weights.h5", "/layers/layer_normalization_63/vars": ["model_00001.weights.h5"], "/layers/dense_62/vars": ["model_00001.weights.h5"], "/layers/dense_63/vars": ["model_00001.weights.h5"], "/layers/layer_scale_63/vars": "model_00001.weights.h5", "/layers/layer_normalization_64/vars": ["model_00001.weights.h5"], "/layers/dense_64/vars": ["model_00001.weights.h5"]}}