| from transformers import PretrainedConfig | |
| class NgwandaConfig(PretrainedConfig): | |
| model_type = "ngwanda" | |
| def __init__(self, d_model=768, attention_heads=8, sequence_length=3072, vocab_size=12000, layers=8, **kwargs): | |
| self.d_model = d_model | |
| self.attention_heads = attention_heads | |
| self.sequence_length = sequence_length | |
| self.vocab_size = vocab_size | |
| self.layers = layers | |
| super().__init__(**kwargs) | |