QLIP-L-14-392 / config.json
zhaoyue-zephyrus's picture
first commit
2bc3344
{
"_name_or_path": "EVA-BSQCLIP",
"architectures": [
"QLIPModel"
],
"auto_map": {
"AutoConfig": "configuration_evaclip.QLIPConfig",
"AutoModel": "modeling_evaclip.QLIPModel"
},
"decoder_config": {
"dropout": 0.0,
"hidden_size": 1024,
"hidden_size_post_q": 1024,
"image_size": 392,
"intermediate_size": 2730,
"k_bias": false,
"layer_norm_eps": 1e-06,
"model_type": "clip_decoder_model",
"num_attention_heads": 16,
"num_hidden_layers": 24,
"patch_size": 14,
"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": 768,
"text_config": {
"bos_token_id": 0,
"dropout": 0.0,
"eos_token_id": 2,
"hidden_size": 768,
"intermediate_size": 3072,
"num_attention_heads": 12,
"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,
"hidden_size": 1024,
"hidden_size_post_q": 1024,
"image_size": 392,
"intermediate_size": 2730,
"k_bias": false,
"layer_norm_eps": 1e-06,
"model_type": "clip_vision_model",
"num_attention_heads": 16,
"num_hidden_layers": 24,
"patch_size": 14,
"projection_dim": 1024,
"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
}