| from transformers import PretrainedConfig | |
| from typing import List | |
| class InfMLLMConfig(PretrainedConfig): | |
| def __init__( | |
| self, | |
| image_size="448", | |
| vit_model="eva_clip_g", | |
| pool_out_size="32", | |
| lm_model="pretrain_models/lmsys/vicuna-7b-v1.5/", | |
| lm_tokenizer="pretrain_models/lmsys/vicuna-7b-v1.5/", | |
| precision="amp_bf16", | |
| **kwargs | |
| ): | |
| self.image_size = image_size | |
| self.vit_model = vit_model | |
| self.pool_out_size = pool_out_size | |
| self.lm_model = lm_model | |
| self.lm_tokenizer = lm_tokenizer | |
| self.precision = precision | |
| super().__init__(**kwargs) | |