Any-to-Any
Safetensors
ml-4m
4M-21_XL / config.json
roman-bachmann's picture
Init
caedbce
raw
history blame
1.12 kB
{
"act_layer": "SiLU",
"decoder_depth": 24,
"dim": 2048,
"domains_in": [
"caption",
"t5_caption",
"det",
"metadata",
"human_poses",
"color_palette",
"sam_instance",
"rgb@224",
"tok_rgb@224",
"tok_normal@224",
"tok_depth@224",
"tok_semseg@224",
"tok_clip@224",
"tok_dinov2@224",
"tok_dinov2_global",
"tok_imagebind@224",
"tok_imagebind_global",
"tok_sam_edge@224",
"tok_canny_edge@224"
],
"domains_out": [
"caption",
"t5_caption",
"det",
"metadata",
"human_poses",
"color_palette",
"sam_instance",
"tok_rgb@224",
"tok_normal@224",
"tok_depth@224",
"tok_semseg@224",
"tok_clip@224",
"tok_dinov2@224",
"tok_dinov2_global",
"tok_imagebind@224",
"tok_imagebind_global",
"tok_sam_edge@224",
"tok_canny_edge@224"
],
"encoder_depth": 24,
"gated_mlp": true,
"image_size": 224,
"mlp_bias": false,
"mlp_ratio": 4,
"norm_bias": false,
"num_heads": 32,
"patch_size": 16,
"proj_bias": false,
"qkv_bias": false,
"share_modality_embeddings": false
}