core.distributed.fsdp.src.megatron_fsdp.experimental#
Experimental Megatron-FSDP implementation.
Submodules#
core.distributed.fsdp.src.megatron_fsdp.experimental.dbuffercore.distributed.fsdp.src.megatron_fsdp.experimental.uneven_dtensorcore.distributed.fsdp.src.megatron_fsdp.experimental.optimizercore.distributed.fsdp.src.megatron_fsdp.experimental.modulecore.distributed.fsdp.src.megatron_fsdp.experimental.countdowncore.distributed.fsdp.src.megatron_fsdp.experimental.quantized_dbuffercore.distributed.fsdp.src.megatron_fsdp.experimental.owner_planningcore.distributed.fsdp.src.megatron_fsdp.experimental.parameter_groupcore.distributed.fsdp.src.megatron_fsdp.experimental.schedulecore.distributed.fsdp.src.megatron_fsdp.experimental.module_utilscore.distributed.fsdp.src.megatron_fsdp.experimental.placementcore.distributed.fsdp.src.megatron_fsdp.experimental.checkpointcore.distributed.fsdp.src.megatron_fsdp.experimental.layoutcore.distributed.fsdp.src.megatron_fsdp.experimental.fully_shardcore.distributed.fsdp.src.megatron_fsdp.experimental.indexed_order
Package Contents#
Data#
API#
- core.distributed.fsdp.src.megatron_fsdp.experimental.__all__#
[‘DBuffer’, ‘Placements’, ‘SchedulePolicy’, ‘fully_shard’, ‘fully_shard_context’, ‘fully_shard_optim…