`bridge.recipes.nemotronh.nemotron_3_nano`#

Module Contents#

Functions#

`nemotron_3_nano_pretrain_config`	Return a pre-training config for Nemotron 3 Nano (30B-A3B MoE).
`nemotron_3_nano_sft_config`	Return a full SFT config for Nemotron 3 Nano (30B-A3B MoE).
`nemotron_3_nano_peft_config`	Return a PEFT config for Nemotron 3 Nano (30B-A3B MoE).

Data#

__all__

API#

bridge.recipes.nemotronh.nemotron_3_nano.nemotron_3_nano_pretrain_config() → megatron.bridge.training.config.ConfigContainer#

Return a pre-training config for Nemotron 3 Nano (30B-A3B MoE).

This is a MoE (Mixture of Experts) model with the following default parallelism:

TP=4, PP=1, EP=8, SP=True
DeepEP enabled for MoE token dispatch

Returns:: Pre-training configuration for Nemotron 3 Nano.
Return type:: ConfigContainer

bridge.recipes.nemotronh.nemotron_3_nano.nemotron_3_nano_sft_config() → megatron.bridge.training.config.ConfigContainer#

Return a full SFT config for Nemotron 3 Nano (30B-A3B MoE).

Default parallelism: TP=1, PP=1, EP=8, SP=False

Returns:: ConfigContainer with all settings pre-configured for Nemotron 3 Nano SFT.

bridge.recipes.nemotronh.nemotron_3_nano.nemotron_3_nano_peft_config( peft_scheme: str | megatron.bridge.peft.base.PEFT = 'lora', ) → megatron.bridge.training.config.ConfigContainer#

Return a PEFT config for Nemotron 3 Nano (30B-A3B MoE).