Motif-Technologies/Motif-Vision-Encoder

1{2  "architectures": [3    "MotifVisionModel"4  ],5  "auto_map": {6    "AutoConfig": "modeling_motif_vision_encoder.MotifVisionConfig",7    "AutoModel": "modeling_motif_vision_encoder.MotifVisionModel"8  },9  "depth": 40,10  "drop_path_rate": 0.0,11  "dtype": "bfloat16",12  "embed_dim": 4096,13  "ffn_bias": true,14  "ffn_layer": "swiglu64",15  "ffn_ratio": 3.0,16  "gated_attention": "elementwise",17  "img_size": 512,18  "in_chans": 3,19  "layerscale_init": 1e-05,20  "mask_k_bias": true,21  "model_type": "motif_vision",22  "n_storage_tokens": 4,23  "norm_layer": "layernormbf16",24  "num_frames": 1,25  "num_heads": 32,26  "patch_size": 16,27  "pos_embed_rope_base": 100.0,28  "pos_embed_rope_rescale_coords": 2.0,29  "proj_bias": true,30  "qk_norm": true,31  "qkv_bias": false,32  "transformers_version": "5.8.1",33  "tubelet_size": 2,34  "untie_cls_and_patch_norms": false,35  "untie_global_and_local_cls_norm": true36}37