| """GR-Lite model configuration.""" | |
| from transformers import PretrainedConfig | |
| class GRLiteConfig(PretrainedConfig): | |
| model_type = "gr_lite" | |
| def __init__( | |
| self, | |
| hidden_size=1024, | |
| num_hidden_layers=24, | |
| num_attention_heads=16, | |
| intermediate_size=4096, | |
| patch_size=16, | |
| image_size=336, | |
| num_channels=3, | |
| num_register_tokens=4, | |
| layer_norm_eps=1e-6, | |
| qkv_bias=True, | |
| k_bias=False, | |
| **kwargs, | |
| ): | |
| super().__init__(**kwargs) | |
| self.hidden_size = hidden_size | |
| self.num_hidden_layers = num_hidden_layers | |
| self.num_attention_heads = num_attention_heads | |
| self.intermediate_size = intermediate_size | |
| self.patch_size = patch_size | |
| self.image_size = image_size | |
| self.num_channels = num_channels | |
| self.num_register_tokens = num_register_tokens | |
| self.layer_norm_eps = layer_norm_eps | |
| self.qkv_bias = qkv_bias | |
| self.k_bias = k_bias | |