`bridge.perf_recipes.llama.vrnvl72.llama3`#

VRNVL72 performance recipes for Llama 3.

Module Contents#

Functions#

`llama3_8b_pretrain_8gpu_vrnvl72_bf16_config`	LLaMA 3 8B pretrain: 8× VRNVL72, BF16 (alias of GB300).
`llama3_8b_pretrain_8gpu_vrnvl72_fp8cs_config`	LLaMA 3 8B pretrain: 8× VRNVL72, FP8-CS (alias of GB300).
`llama3_8b_pretrain_8gpu_vrnvl72_fp8mx_config`	LLaMA 3 8B pretrain: 8× VRNVL72, FP8-MX (alias of GB300).
`llama3_8b_pretrain_8gpu_vrnvl72_nvfp4_config`	LLaMA 3 8B pretrain: 8× VRNVL72, NVFP4 (alias of GB300).
`llama3_70b_pretrain_64gpu_vrnvl72_bf16_config`	LLaMA 3 70B pretrain: 64× VRNVL72, BF16 (alias of GB300).
`llama3_70b_pretrain_64gpu_vrnvl72_fp8mx_config`	LLaMA 3 70B pretrain: 64× VRNVL72, FP8-MX (alias of GB300).
`llama3_70b_pretrain_64gpu_vrnvl72_nvfp4_config`	LLaMA 3 70B pretrain: 64× VRNVL72, NVFP4 (alias of GB300).

API#

bridge.perf_recipes.llama.vrnvl72.llama3.llama3_8b_pretrain_8gpu_vrnvl72_bf16_config() → megatron.bridge.perf_recipes.llama.common.ConfigContainer#: LLaMA 3 8B pretrain: 8× VRNVL72, BF16 (alias of GB300).

bridge.perf_recipes.llama.vrnvl72.llama3.llama3_8b_pretrain_8gpu_vrnvl72_fp8cs_config() → megatron.bridge.perf_recipes.llama.common.ConfigContainer#: LLaMA 3 8B pretrain: 8× VRNVL72, FP8-CS (alias of GB300).

bridge.perf_recipes.llama.vrnvl72.llama3.llama3_8b_pretrain_8gpu_vrnvl72_fp8mx_config() → megatron.bridge.perf_recipes.llama.common.ConfigContainer#: LLaMA 3 8B pretrain: 8× VRNVL72, FP8-MX (alias of GB300).

bridge.perf_recipes.llama.vrnvl72.llama3.llama3_8b_pretrain_8gpu_vrnvl72_nvfp4_config() → megatron.bridge.perf_recipes.llama.common.ConfigContainer#: LLaMA 3 8B pretrain: 8× VRNVL72, NVFP4 (alias of GB300).

bridge.perf_recipes.llama.vrnvl72.llama3.llama3_70b_pretrain_64gpu_vrnvl72_bf16_config() → megatron.bridge.perf_recipes.llama.common.ConfigContainer#: LLaMA 3 70B pretrain: 64× VRNVL72, BF16 (alias of GB300).

bridge.perf_recipes.llama.vrnvl72.llama3.llama3_70b_pretrain_64gpu_vrnvl72_fp8mx_config() → megatron.bridge.perf_recipes.llama.common.ConfigContainer#: LLaMA 3 70B pretrain: 64× VRNVL72, FP8-MX (alias of GB300).

bridge.perf_recipes.llama.vrnvl72.llama3.llama3_70b_pretrain_64gpu_vrnvl72_nvfp4_config() → megatron.bridge.perf_recipes.llama.common.ConfigContainer#: LLaMA 3 70B pretrain: 64× VRNVL72, NVFP4 (alias of GB300).