bridge.models.sarvam.sarvam_moe_bridge#

Module Contents#

Classes#

SarvamMoEBridge

Megatron Hub Bridge for Sarvam MoE Causal LM.

API#

class bridge.models.sarvam.sarvam_moe_bridge.SarvamMoEBridge#

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

Megatron Hub Bridge for Sarvam MoE Causal LM.

This bridge handles the conversion between HuggingFace SarvamMoEForCausalLM and Megatron-Core GPTModel formats. Sarvam MoE models use mixture of experts architecture with QKV layernorm.

provider_bridge(
hf_pretrained: megatron.bridge.models.hf_pretrained.causal_lm.PreTrainedCausalLM,
) megatron.bridge.models.sarvam.sarvam_provider.SarvamMoEModelProvider#
mapping_registry() megatron.bridge.models.conversion.mapping_registry.MegatronMappingRegistry#