`bridge.perf_recipes.llama.vrnvl72.llama31`#

VRNVL72 performance recipes for Llama 3.1.

Module Contents#

Functions#

`llama31_405b_pretrain_256gpu_vrnvl72_bf16_config`	Llama3.1 405B pretrain: 256x VRNVL72, BF16 (alias of GB300).
`llama31_405b_pretrain_256gpu_vrnvl72_fp8mx_config`	Llama3.1 405B pretrain: 256x VRNVL72, FP8-MX (alias of GB300).
`llama31_405b_pretrain_256gpu_vrnvl72_nvfp4_config`	Llama3.1 405B pretrain: 256x VRNVL72, NVFP4 (alias of GB300).

API#

bridge.perf_recipes.llama.vrnvl72.llama31.llama31_405b_pretrain_256gpu_vrnvl72_bf16_config() → megatron.bridge.perf_recipes.llama.common.ConfigContainer#: Llama3.1 405B pretrain: 256x VRNVL72, BF16 (alias of GB300).

bridge.perf_recipes.llama.vrnvl72.llama31.llama31_405b_pretrain_256gpu_vrnvl72_fp8mx_config() → megatron.bridge.perf_recipes.llama.common.ConfigContainer#: Llama3.1 405B pretrain: 256x VRNVL72, FP8-MX (alias of GB300).

bridge.perf_recipes.llama.vrnvl72.llama31.llama31_405b_pretrain_256gpu_vrnvl72_nvfp4_config() → megatron.bridge.perf_recipes.llama.common.ConfigContainer#: Llama3.1 405B pretrain: 256x VRNVL72, NVFP4 (alias of GB300).