`bridge.recipes.nemotronh.h100.nemotron_3_nano`#

Module Contents#

Functions#

`_nemotron_3_nano_finetune_model`	Build the Nemotron 3 Nano finetuning provider.
`nemotron_3_nano_pretrain_8gpu_h100_bf16_config`	Return a pre-training config for Nemotron 3 Nano (30B-A3B MoE).
`nemotron_3_5_lightning_pretrain_config`	Return the Nemotron 3.5 Lightning BF16 pretraining config.
`nemotron_3_nano_sft_8gpu_h100_bf16_config`	Return a full SFT config for Nemotron 3 Nano.
`nemotron_3_5_lightning_sft_config`	Return a full SFT config for Nemotron 3.5 Lightning.
`nemotron_3_5_lightning_sft_openmathinstruct2_packed_config`	Return the verified 4K packed OpenMathInstruct-2 SFT config.
`_nemotron_3_nano_peft_8gpu_h100_bf16_config`	Build a PEFT config for Nemotron 3 Nano.
`nemotron_3_nano_peft_8gpu_h100_bf16_config`	Return a PEFT config for Nemotron 3 Nano.
`nemotron_3_5_lightning_peft_config`	Return a PEFT config for Nemotron 3.5 Lightning.

Data#

`_NEMOTRON_3_NANO_MODEL_ID`
`_NEMOTRON_3_5_LIGHTNING_MODEL_ID`
`_NEMOTRON_3_5_LIGHTNING_MODEL_REVISION`
`_OPENMATHINSTRUCT2_REVISION`
`__all__`

API#

bridge.recipes.nemotronh.h100.nemotron_3_nano._NEMOTRON_3_NANO_MODEL_ID#: ‘nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16’

bridge.recipes.nemotronh.h100.nemotron_3_nano._NEMOTRON_3_5_LIGHTNING_MODEL_ID#: ‘nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16’

bridge.recipes.nemotronh.h100.nemotron_3_nano._NEMOTRON_3_5_LIGHTNING_MODEL_REVISION#: ‘b3caaabed0263651a17dc1f2d4ce97e794f76c44’

bridge.recipes.nemotronh.h100.nemotron_3_nano._OPENMATHINSTRUCT2_REVISION#: ‘469216e3f46f4dacf476b382e192485ea51a143e’

bridge.recipes.nemotronh.h100.nemotron_3_nano._nemotron_3_nano_finetune_model() → megatron.bridge.models.hybrid.hybrid_provider.HybridModelProvider#: Build the Nemotron 3 Nano finetuning provider.

bridge.recipes.nemotronh.h100.nemotron_3_nano.nemotron_3_nano_pretrain_8gpu_h100_bf16_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=1, PP=1, ETP=1, EP=8, SP=False
HybridEP enabled for MoE token dispatch

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

bridge.recipes.nemotronh.h100.nemotron_3_nano.nemotron_3_5_lightning_pretrain_config() → megatron.bridge.training.config.ConfigContainer#: Return the Nemotron 3.5 Lightning BF16 pretraining config.

bridge.recipes.nemotronh.h100.nemotron_3_nano.nemotron_3_nano_sft_8gpu_h100_bf16_config() → megatron.bridge.training.config.ConfigContainer#

Return a full SFT config for Nemotron 3 Nano.

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

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

bridge.recipes.nemotronh.h100.nemotron_3_nano.nemotron_3_5_lightning_sft_config() → megatron.bridge.training.config.ConfigContainer#: Return a full SFT config for Nemotron 3.5 Lightning.

bridge.recipes.nemotronh.h100.nemotron_3_nano.nemotron_3_5_lightning_sft_openmathinstruct2_packed_config() → megatron.bridge.training.config.ConfigContainer#: Return the verified 4K packed OpenMathInstruct-2 SFT config.

bridge.recipes.nemotronh.h100.nemotron_3_nano._nemotron_3_nano_peft_8gpu_h100_bf16_config( peft_scheme: str | megatron.bridge.peft.base.PEFT = 'lora', ) → megatron.bridge.training.config.ConfigContainer#

Build a PEFT config for Nemotron 3 Nano.