vision_tower/config.json

{
  "_name_or_path": "",
  "architectures": [
    "RQVAESIGLIPTransformer"
  ],
  "hidden_size": 1024,
  "model_type": "rqvaesigliptransformer_model",
  "rqtransformer": {
    "architectures": [
      "RQTransformer"
    ],
    "block_size": [
      16,
      16,
      4
    ],
    "embed_dim": 2560,
    "head": {
      "block": {
        "n_head": 40
      },
      "n_layer": 6
    },
    "input_embed_dim_1": 1024,
    "input_embed_dim_2": 4096,
    "model_type": "rqtransformer_model",
    "torch_dtype": "float32",
    "transformers_version": "4.36.2",
    "vocab_size": 16384
  }
  ,
  "rqvaesiglip": {
    "architectures": [
      "RQVAESiglip"
    ],
    "bottleneck_type": "rq",
    "checkpointing": true,
    "ckpt_path": null,
    "code_shape": [
      16,
      16,
      4
    ],
    "ddconfig": {
      "attn_resolutions": [
        16
      ],
      "ch": 128,
      "ch_mult": [
        1,
        1,
        2,
        2,
        4
      ],
      "double_z": false,
      "dropout": 0.0,
      "in_channels": 3,
      "num_res_blocks": 2,
      "out_ch": 3,
      "resolution": 256,
      "z_channels": 256
    },
    "decay": 0.99,
    "embed_dim": 1024,
    "hidden_size": 1024,
    "ignore_keys": null,
    "latent_loss_weight": 0.25,
    "latent_shape": [
      16,
      16,
      1024
    ],
    "loss_type": "mse",
    "model_type": "rqvaesiglip_model",
    "n_embed": 16384,
    "pretrained_model": "google/siglip-large-patch16-256",
    "restart_unused_codes": true,
    "shared_codebook": true,
    "torch_dtype": "float32",
    "transformers_version": "4.36.2"
  }
  ,
  "torch_dtype": "bfloat16",
  "transformers_version": "4.36.2"
}