bridge.recipes.deepseek.gb200.deepseek_v4#

GB200 recipes for DeepSeek V4 Flash.

Validated on OCI GB200 NVL72 (64 GPUs, PP=8, EP=8).

Module Contents#

Functions#

deepseek_v4_flash_pretrain_64gpu_gb200_bf16_config

DeepSeek-V4-Flash BF16 pre-training on 64-GPU GB200 (PP=8, EP=8).

deepseek_v4_flash_pretrain_64gpu_gb200_fp8mx_config

DeepSeek-V4-Flash MXFP8 pre-training on 64-GPU GB200 (PP=8, EP=8).

deepseek_v4_flash_pretrain_64gpu_gb200_bf16_muon_config

DeepSeek-V4-Flash BF16 Muon pre-training on 64-GPU GB200 (PP=8, EP=8).

API#

bridge.recipes.deepseek.gb200.deepseek_v4.deepseek_v4_flash_pretrain_64gpu_gb200_bf16_config() megatron.bridge.training.config.ConfigContainer#

DeepSeek-V4-Flash BF16 pre-training on 64-GPU GB200 (PP=8, EP=8).

All features validated in 1k-step convergence runs on OCI GB200 NVL72: fused DSA indexer, DSA indexer loss, HybridEP dispatcher, GroupedGEMM, permute fusion, selective recompute with mla_up_proj.

Note: set checkpoint.save_optim=False when using HybridEP.

bridge.recipes.deepseek.gb200.deepseek_v4.deepseek_v4_flash_pretrain_64gpu_gb200_fp8mx_config() megatron.bridge.training.config.ConfigContainer#

DeepSeek-V4-Flash MXFP8 pre-training on 64-GPU GB200 (PP=8, EP=8).

Note: set checkpoint.save_optim=False when using HybridEP.

bridge.recipes.deepseek.gb200.deepseek_v4.deepseek_v4_flash_pretrain_64gpu_gb200_bf16_muon_config() megatron.bridge.training.config.ConfigContainer#

DeepSeek-V4-Flash BF16 Muon pre-training on 64-GPU GB200 (PP=8, EP=8).