bridge.recipes.bagel.h100.bagel#

BAGEL-7B-MoT training recipes.

Module Contents#

Functions#

bagel_7b_pretrain_8gpu_h100_bf16_config

Return an 8-GPU Megatron-FSDP BAGEL pretraining configuration.

bagel_7b_pretrain_32gpu_h100_bf16_config

Return a 32-GPU pure-data-parallel Megatron-FSDP BAGEL configuration.

bagel_7b_finetune_8gpu_h100_bf16_config

Return the official BAGEL fine-tuning overrides for the 8-GPU recipe.

Data#

API#

bridge.recipes.bagel.h100.bagel.bagel_7b_pretrain_8gpu_h100_bf16_config() → megatron.bridge.training.config.ConfigContainer#

Return an 8-GPU Megatron-FSDP BAGEL pretraining configuration.

bridge.recipes.bagel.h100.bagel.bagel_7b_pretrain_32gpu_h100_bf16_config() → megatron.bridge.training.config.ConfigContainer#

Return a 32-GPU pure-data-parallel Megatron-FSDP BAGEL configuration.

bridge.recipes.bagel.h100.bagel.bagel_7b_finetune_8gpu_h100_bf16_config() → megatron.bridge.training.config.ConfigContainer#

Return the official BAGEL fine-tuning overrides for the 8-GPU recipe.

bridge.recipes.bagel.h100.bagel.__all__#

[‘bagel_7b_finetune_8gpu_h100_bf16_config’, ‘bagel_7b_pretrain_32gpu_h100_bf16_config’, ‘bagel_7b_pr…