lam/model_config.json

{
    "architecture": "IgorModel",
    "description": "",
    "version": 0,
    "config": {
        "resolution": 224,
        "patch_size": 14,
        "in_channels": 3,
        "d_t": 8,
        "augment_type": "resize_crop",
        "augment_level": "clip",
        "random_crop_scale": [
            0.8,
            1.0
        ],
        "random_crop_ratio": [
            0.9,
            1.1
        ],
        "mlp_ratio": 4.0,
        "use_xformers": null,
        "encoder_depth": 12,
        "encoder_embed_dim": 768,
        "encoder_n_heads": 12,
        "action_latent_dim": 32,
        "st_use_qk_norm": true,
        "num_learned_tokens": 2,
        "map_heads": 24,
        "decoder_depth": 12,
        "decoder_embed_dim": 768,
        "decoder_n_heads": 12,
        "use_qk_norm": true,
        "n_codes": 32,
        "grid_size": 16,
        "embed_tokens": 256
    }
}