{ "architectures": [ "FASTForImageCaptioning" ], "backbone_act_func": "relu", "backbone_bias": false, "backbone_dilation": 1, "backbone_dropout_rate": 0, "backbone_groups": 1, "backbone_has_shuffle": false, "backbone_in_channels": 3, "backbone_kernel_size": 3, "backbone_ops_order": "weight_bn_act", "backbone_out_channels": 64, "backbone_stage1_dilation": [ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 ], "backbone_stage1_groups": [ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 ], "backbone_stage1_in_channels": [ 64, 64, 64, 64, 64, 64, 64, 64, 64, 64 ], "backbone_stage1_kernel_size": [ [ 3, 3 ], [ 3, 3 ], [ 3, 1 ], [ 3, 3 ], [ 3, 1 ], [ 3, 3 ], [ 3, 3 ], [ 1, 3 ], [ 3, 3 ], [ 3, 3 ] ], "backbone_stage1_out_channels": [ 64, 64, 64, 64, 64, 64, 64, 64, 64, 64 ], "backbone_stage1_stride": [ 1, 2, 1, 1, 1, 1, 1, 1, 1, 1 ], "backbone_stage2_dilation": [ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 ], "backbone_stage2_groups": [ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 ], "backbone_stage2_in_channels": [ 64, 128, 128, 128, 128, 128, 128, 128, 128, 128 ], "backbone_stage2_kernel_size": [ [ 3, 3 ], [ 1, 3 ], [ 3, 3 ], [ 3, 1 ], [ 3, 3 ], [ 3, 3 ], [ 3, 1 ], [ 3, 1 ], [ 3, 3 ], [ 3, 3 ] ], "backbone_stage2_out_channels": [ 128, 128, 128, 128, 128, 128, 128, 128, 128, 128 ], "backbone_stage2_stride": [ 2, 1, 1, 1, 1, 1, 1, 1, 1, 1 ], "backbone_stage3_dilation": [ 1, 1, 1, 1, 1, 1, 1, 1 ], "backbone_stage3_groups": [ 1, 1, 1, 1, 1, 1, 1, 1 ], "backbone_stage3_in_channels": [ 128, 256, 256, 256, 256, 256, 256, 256 ], "backbone_stage3_kernel_size": [ [ 3, 3 ], [ 3, 3 ], [ 3, 3 ], [ 1, 3 ], [ 3, 3 ], [ 3, 1 ], [ 3, 3 ], [ 3, 1 ] ], "backbone_stage3_out_channels": [ 256, 256, 256, 256, 256, 256, 256, 256 ], "backbone_stage3_stride": [ 2, 1, 1, 1, 1, 1, 1, 1 ], "backbone_stage4_dilation": [ 1, 1, 1, 1, 1 ], "backbone_stage4_groups": [ 1, 1, 1, 1, 1 ], "backbone_stage4_in_channels": [ 256, 512, 512, 512, 512 ], "backbone_stage4_kernel_size": [ [ 3, 3 ], [ 1, 3 ], [ 3, 1 ], [ 3, 1 ], [ 1, 3 ] ], "backbone_stage4_out_channels": [ 512, 512, 512, 512, 512 ], "backbone_stage4_stride": [ 2, 1, 1, 1, 1 ], "backbone_stride": 2, "backbone_use_bn": true, "bbox_type": 300, "head_conv_dilation": 1, "head_conv_groups": 1, "head_conv_in_channels": 512, "head_conv_kernel_size": [ 3, 3 ], "head_conv_out_channels": 128, "head_conv_stride": 1, "head_dropout_ratio": 0.1, "head_final_act_func": null, "head_final_bias": false, "head_final_dilation": 1, "head_final_dropout_rate": 0, "head_final_groups": 1, "head_final_has_shuffle": false, "head_final_in_channels": 128, "head_final_kernel_size": 1, "head_final_ops_order": "weight", "head_final_out_channels": 5, "head_final_stride": 1, "head_final_use_bn": false, "head_pooling_size": 11, "initializer_range": 0.02, "loss_bg": false, "min_area": 300, "min_score": 300, "neck_dilation": [ 1, 1, 1, 1 ], "neck_groups": [ 1, 1, 1, 1 ], "neck_in_channels": [ 64, 128, 256, 512 ], "neck_kernel_size": [ [ 3, 3 ], [ 3, 3 ], [ 3, 3 ], [ 3, 3 ] ], "neck_out_channels": [ 128, 128, 128, 128 ], "neck_stride": [ 1, 1, 1, 1 ], "torch_dtype": "float32", "transformers_version": "4.34.0.dev0" }