bridge.models.deepseek.deepseek_v3_bridge

Module Contents

Classes

DeepSeekV3Bridge

Megatron Bridge for DeepSeek-V3.

API

class bridge.models.deepseek.deepseek_v3_bridge.DeepSeekV3Bridge

Bases: megatron.bridge.models.conversion.model_bridge.MegatronModelBridge

Megatron Bridge for DeepSeek-V3.

As a user you would not use this bridge directly, but through AutoBridge.

.. rubric:: Example

from megatron.bridge import AutoBridge bridge = AutoBridge.from_hf_pretrained(“deepseek-ai/DeepSeek-V3-Base”, trust_remote_code=True) provider = bridge.to_megatron_provider()

provider_bridge(

hf_pretrained: megatron.bridge.models.hf_pretrained.causal_lm.PreTrainedCausalLM,

) → megatron.bridge.models.deepseek.deepseek_provider.DeepSeekV3ModelProvider

mapping_registry() → megatron.bridge.models.conversion.mapping_registry.MegatronMappingRegistry

maybe_modify_converted_hf_weight(

task: megatron.bridge.models.conversion.model_bridge.WeightConversionTask,

converted_weights_dict: Dict[str, torch.Tensor],

hf_state_dict: Mapping[str, torch.Tensor],

) → Dict[str, torch.Tensor]

Add rotary embedding inverse frequency parameter if needed but not present. This is needed for moonshotai related models (e.g., Moonlight-16B-A3B-Instruct).