maxtext.checkpoint_conversion.utils.hf_model_configs module

maxtext.checkpoint_conversion.utils.hf_model_configs module#

This config defines the architectural configurations of the Hugging Face version of a model.

class maxtext.checkpoint_conversion.utils.hf_model_configs.DeepseekV32Config(**kwargs)[source]#

Bases: PreTrainedConfig

model_type: ClassVar[str] = 'deepseek_v32'#