config.json

{
  "_name_or_path": "EVA-BSQCLIP",
  "architectures": [
    "QLIPModel"
  ],
  "auto_map": {
    "AutoConfig": "configuration_evaclip.QLIPConfig",
    "AutoModel": "modeling_evaclip.QLIPModel"
  },
  "decoder_config": {
    "dropout": 0.0,
    "image_size": 256,
    "intermediate_size": 2048,
    "k_bias": false,
    "layer_norm_eps": 1e-06,
    "model_type": "clip_decoder_model",
    "patch_size": 16,
    "rope": true,
    "rope_shift": 0,
    "subln": true,
    "swiglu": true,
    "use_bfloat16": true,
    "use_rms_norm": true
  },
  "initializer_factor": 1.0,
  "logit_scale_init_value": 2.6592,
  "model_type": "clip",
  "projection_dim": 512,
  "text_config": {
    "bos_token_id": 0,
    "dropout": 0.0,
    "eos_token_id": 2,
    "model_type": "clip_text_model",
    "use_bfloat16": true,
    "use_rms_norm": false
  },
  "text_projection_bias": false,
  "torch_dtype": "float32",
  "transformers_version": "4.37.2",
  "vision_config": {
    "dropout": 0.0,
    "image_size": 256,
    "intermediate_size": 2048,
    "k_bias": false,
    "layer_norm_eps": 1e-06,
    "model_type": "clip_vision_model",
    "patch_size": 16,
    "quantizer": "bsq",
    "quantizer_cfg": {
      "embed_dim": 28,
      "group_size": 1,
      "input_format": "blc",
      "inv_temperature": 1.0,
      "l2_norm": true
    },
    "quantizer_embed_type": "mlp",
    "quantizer_l2_norm": true,
    "rope": true,
    "rope_shift": 1,
    "subln": true,
    "swiglu": true,
    "use_bfloat16": true,
    "use_rms_norm": true
  },
  "vision_projection_bias": true
}