bridge.perf_recipes.nemotronh.gb300.nemotronh

GB300 performance recipes for NemotronH and Nemotron 3.

Module Contents

Functions

nemotronh_56b_pretrain_64gpu_gb300_fp8cs_config	NemotronH 56B pretrain: 64× GB300, FP8 current-scaling.
nemotron_3_super_pretrain_64gpu_gb300_bf16_config	Nemotron 3 Super pretrain: 64× GB300, BF16.
nemotron_3_super_pretrain_64gpu_gb300_fp8mx_config	Nemotron 3 Super pretrain: 64× GB300, MXFP8.
nemotron_3_super_pretrain_64gpu_gb300_nvfp4_config	Nemotron 3 Super pretrain: 64× GB300, NVFP4.
_nemotron_3_ultra_gb300_fp8mx_config	Shared builder for Nemotron 3 Ultra GB300 MXFP8 Megatron-FSDP perf recipes.
nemotron_3_ultra_pretrain_256gpu_gb300_fp8mx_config	Nemotron 3 Ultra (550B-A55B LatentMoE) pretrain: 256× GB300, MXFP8, Megatron-FSDP (HSDP).
nemotron_3_nano_pretrain_8gpu_gb300_bf16_config	Nemotron 3 Nano pretrain: 8× GB300, BF16.
nemotron_3_nano_pretrain_8gpu_gb300_fp8mx_config	Nemotron 3 Nano pretrain: 8× GB300, MXFP8.
nemotron_3_nano_pretrain_8gpu_gb300_nvfp4_config	Nemotron 3 Nano pretrain: 8× GB300, NVFP4.
nemotronh_56b_pretrain_256gpu_gb300_bf16_config	NemotronH 56B pretrain: 256× GB300, BF16 (same layout as FP8-CS).
nemotronh_56b_pretrain_256gpu_gb300_fp8cs_config	NemotronH 56B pretrain: 256× GB300, FP8 current-scaling, legacy-scaled GBS.

API

bridge.perf_recipes.nemotronh.gb300.nemotronh.nemotronh_56b_pretrain_64gpu_gb300_fp8cs_config() → megatron.bridge.perf_recipes.nemotronh.common.ConfigContainer

NemotronH 56B pretrain: 64× GB300, FP8 current-scaling.

bridge.perf_recipes.nemotronh.gb300.nemotronh.nemotron_3_super_pretrain_64gpu_gb300_bf16_config() → megatron.bridge.perf_recipes.nemotronh.common.ConfigContainer

Nemotron 3 Super pretrain: 64× GB300, BF16.

bridge.perf_recipes.nemotronh.gb300.nemotronh.nemotron_3_super_pretrain_64gpu_gb300_fp8mx_config() → megatron.bridge.perf_recipes.nemotronh.common.ConfigContainer

Nemotron 3 Super pretrain: 64× GB300, MXFP8.

bridge.perf_recipes.nemotronh.gb300.nemotronh.nemotron_3_super_pretrain_64gpu_gb300_nvfp4_config() → megatron.bridge.perf_recipes.nemotronh.common.ConfigContainer

Nemotron 3 Super pretrain: 64× GB300, NVFP4.

bridge.perf_recipes.nemotronh.gb300.nemotronh._nemotron_3_ultra_gb300_fp8mx_config(

*,

num_gpus: int,

expert_model_parallel_size: int,

global_batch_size: int,

) → megatron.bridge.perf_recipes.nemotronh.common.ConfigContainer

Shared builder for Nemotron 3 Ultra GB300 MXFP8 Megatron-FSDP perf recipes.

bridge.perf_recipes.nemotronh.gb300.nemotronh.nemotron_3_ultra_pretrain_256gpu_gb300_fp8mx_config() → megatron.bridge.perf_recipes.nemotronh.common.ConfigContainer

Nemotron 3 Ultra (550B-A55B LatentMoE) pretrain: 256× GB300, MXFP8, Megatron-FSDP (HSDP).

TP1 / PP1 / CP1 / EP64 / ETP1, GBS 256 / MBS 1, seq 8192, BF16 + MXFP8 mixed precision, HybridEP flex dispatcher, CuteDSL fused grouped MLP, selective recompute + fine-grained activation offload of the expert MLP, MTP=2.

bridge.perf_recipes.nemotronh.gb300.nemotronh.nemotron_3_nano_pretrain_8gpu_gb300_bf16_config() → megatron.bridge.perf_recipes.nemotronh.common.ConfigContainer

Nemotron 3 Nano pretrain: 8× GB300, BF16.

bridge.perf_recipes.nemotronh.gb300.nemotronh.nemotron_3_nano_pretrain_8gpu_gb300_fp8mx_config() → megatron.bridge.perf_recipes.nemotronh.common.ConfigContainer

Nemotron 3 Nano pretrain: 8× GB300, MXFP8.

bridge.perf_recipes.nemotronh.gb300.nemotronh.nemotron_3_nano_pretrain_8gpu_gb300_nvfp4_config() → megatron.bridge.perf_recipes.nemotronh.common.ConfigContainer

Nemotron 3 Nano pretrain: 8× GB300, NVFP4.

bridge.perf_recipes.nemotronh.gb300.nemotronh.nemotronh_56b_pretrain_256gpu_gb300_bf16_config() → megatron.bridge.perf_recipes.nemotronh.common.ConfigContainer

NemotronH 56B pretrain: 256× GB300, BF16 (same layout as FP8-CS).

bridge.perf_recipes.nemotronh.gb300.nemotronh.nemotronh_56b_pretrain_256gpu_gb300_fp8cs_config() → megatron.bridge.perf_recipes.nemotronh.common.ConfigContainer

NemotronH 56B pretrain: 256× GB300, FP8 current-scaling, legacy-scaled GBS.