{ "model_type": "fc_clip", "auto_map": { "AutoConfig": "configuration_fc_clip.FCCLIPConfig", "AutoModel": "modeling_fc_clip.FCCLIPForPanopticSegmentation" }, "clip_model_name": "convnext_large_d_320", "clip_pretrained": "laion2b_s29b_b131k_ft_soup", "clip_embedding_dim": 768, "conv_dim": 256, "mask_dim": 256, "transformer_dropout": 0.0, "transformer_nheads": 8, "transformer_dim_feedforward": 1024, "transformer_enc_layers": 6, "transformer_in_features": ["res3", "res4", "res5"], "in_features": ["res2", "res3", "res4", "res5"], "common_stride": 4, "hidden_dim": 256, "num_queries": 250, "nheads": 8, "dim_feedforward": 2048, "dec_layers": 9, "pre_norm": false, "enforce_input_project": false, "num_classes": 133, "num_thing_classes": 80, "geometric_ensemble_alpha": 0.4, "geometric_ensemble_beta": 0.8, "ensemble_on_valid_mask": false, "object_mask_threshold": 0.8, "overlap_threshold": 0.8, "pixel_mean": [122.7709383, 116.7460125, 104.09373615], "pixel_std": [68.5005327, 66.6321579, 70.32316305] }