Index

A

abort() (torchft.process_group.ProcessGroup method) (torchft.process_group.ProcessGroupNCCL method) (torchft.process_group.ProcessGroupWrapper method) add_param_group() (torchft.optim.OptimizerWrapper method) address (torchft.coordination.QuorumMember attribute) address() (torchft.checkpointing.HTTPTransport method) (torchft.coordination.LighthouseServer method) (torchft.coordination.ManagerServer method) (torchft.parameter_server.ParameterServer method) allgather() (torchft.process_group.ProcessGroup method) (torchft.process_group.ProcessGroupBaby method) (torchft.process_group.ProcessGroupDummy method) (torchft.process_group.ProcessGroupWrapper method) allgather_into_tensor_coalesced() (torchft.process_group.ProcessGroup method) (torchft.process_group.ProcessGroupBaby method) (torchft.process_group.ProcessGroupDummy method) (torchft.process_group.ProcessGroupWrapper method)

allow_checkpoint() (torchft.checkpointing.HTTPTransport method) allreduce() (torchft.manager.Manager method) (torchft.process_group.ErrorSwallowingProcessGroupWrapper method) (torchft.process_group.FakeProcessGroupWrapper method) (torchft.process_group.ManagedProcessGroup method) (torchft.process_group.ProcessGroup method) (torchft.process_group.ProcessGroupBaby method) (torchft.process_group.ProcessGroupDummy method) (torchft.process_group.ProcessGroupWrapper method) allreduce_coalesced() (torchft.process_group.ProcessGroup method) (torchft.process_group.ProcessGroupBaby method) (torchft.process_group.ProcessGroupDummy method) (torchft.process_group.ProcessGroupWrapper method) alltoall_base() (torchft.process_group.ProcessGroup method) (torchft.process_group.ProcessGroupBaby method) (torchft.process_group.ProcessGroupDummy method) (torchft.process_group.ProcessGroupWrapper method)

B

barrier() (torchft.process_group.ProcessGroup method) (torchft.process_group.ProcessGroupBaby method) (torchft.process_group.ProcessGroupDummy method) (torchft.process_group.ProcessGroupWrapper method) batches_committed() (torchft.manager.Manager method)

broadcast() (torchft.process_group.ProcessGroup method) (torchft.process_group.ProcessGroupBaby method) (torchft.process_group.ProcessGroupDummy method) (torchft.process_group.ProcessGroupWrapper method) broadcast_one() (torchft.process_group.ProcessGroup method)

C

CheckpointTransport (class in torchft.checkpointing) configure() (torchft.process_group.ErrorSwallowingProcessGroupWrapper method) (torchft.process_group.FakeProcessGroupWrapper method) (torchft.process_group.ProcessGroup method) (torchft.process_group.ProcessGroupBaby method) (torchft.process_group.ProcessGroupDummy method) (torchft.process_group.ProcessGroupWrapper method)

create_store_client() (in module torchft.process_group) created (torchft.coordination.Quorum attribute) current_step() (torchft.manager.Manager method)

D

data (torchft.coordination.QuorumMember attribute) DiLoCo (class in torchft.local_sgd) disallow_checkpoint() (torchft.checkpointing.CheckpointTransport method) (torchft.checkpointing.HTTPTransport method)

DistributedDataParallel (class in torchft.ddp) DistributedSampler (class in torchft.data) DYNAMIC (torchft.manager.WorldSizeMode attribute)

E

error() (torchft.process_group.ErrorSwallowingProcessGroupWrapper method) errored() (torchft.manager.Manager method) (torchft.process_group.ProcessGroup method) (torchft.process_group.ProcessGroupNCCL method)

ErrorSwallowingProcessGroupWrapper (class in torchft.process_group) ExceptionWithTraceback extract_local_tensor() (in module torchft.local_sgd)

F

FakeProcessGroupWrapper (class in torchft.process_group) FIXED_WITH_SPARES (torchft.manager.WorldSizeMode attribute)

forward() (torchft.ddp.PureDistributedDataParallel method) (torchft.parameter_server.ParameterServer method)

G

getBackendName() (torchft.process_group.ManagedProcessGroup method) (torchft.process_group.ProcessGroup method) (torchft.process_group.ProcessGroupBabyGloo method) (torchft.process_group.ProcessGroupBabyNCCL method) (torchft.process_group.ProcessGroupDummy method) (torchft.process_group.ProcessGroupGloo method) (torchft.process_group.ProcessGroupNCCL method) (torchft.process_group.ProcessGroupWrapper method)

group_name (torchft.process_group.ProcessGroup property)

H

heartbeat() (torchft.coordination.LighthouseClient method)

HTTPTransport (class in torchft.checkpointing)

I

is_participating() (torchft.manager.Manager method)

L

LighthouseClient (class in torchft.coordination) LighthouseServer (class in torchft.coordination)

load_state_dict() (torchft.manager.Manager method) (torchft.optim.OptimizerWrapper method) LocalSGD (class in torchft.local_sgd)

M

ManagedProcessGroup (class in torchft.process_group) Manager (class in torchft.manager) ManagerClient (class in torchft.coordination) ManagerServer (class in torchft.coordination) metadata() (torchft.checkpointing.CheckpointTransport method) (torchft.checkpointing.HTTPTransport method) module torchft.checkpointing torchft.coordination torchft.data torchft.ddp torchft.local_sgd torchft.manager torchft.optim torchft.parameter_server torchft.process_group

N

new_process_group() (torchft.parameter_server.ParameterServer class method) new_session() (torchft.parameter_server.ParameterServer class method)

num_active_work() (torchft.process_group.ProcessGroupBaby method) num_participants() (torchft.manager.Manager method)

O

OptimizerWrapper (class in torchft.optim)

P

param_groups (torchft.optim.OptimizerWrapper property) ParameterServer (class in torchft.parameter_server) parent (torchft.process_group.ProcessGroupWrapper property) participants (torchft.coordination.Quorum attribute) participating_rank() (torchft.manager.Manager method) ProcessGroup (class in torchft.process_group) ProcessGroupBaby (class in torchft.process_group)

ProcessGroupBabyGloo (class in torchft.process_group) ProcessGroupBabyNCCL (class in torchft.process_group) ProcessGroupDummy (class in torchft.process_group) ProcessGroupGloo (class in torchft.process_group) ProcessGroupNCCL (class in torchft.process_group) ProcessGroupWrapper (class in torchft.process_group) PureDistributedDataParallel (class in torchft.ddp)

Q

Quorum (class in torchft.coordination) quorum() (torchft.coordination.LighthouseClient method)

quorum_id (torchft.coordination.Quorum attribute) QuorumMember (class in torchft.coordination)

R

recv() (torchft.process_group.ProcessGroup method) (torchft.process_group.ProcessGroupBaby method) (torchft.process_group.ProcessGroupDummy method) (torchft.process_group.ProcessGroupWrapper method) recv_checkpoint() (torchft.checkpointing.CheckpointTransport method) (torchft.checkpointing.HTTPTransport method) reduce_scatter() (torchft.process_group.ProcessGroup method) (torchft.process_group.ProcessGroupBaby method) (torchft.process_group.ProcessGroupBabyGloo method) (torchft.process_group.ProcessGroupDummy method) (torchft.process_group.ProcessGroupGloo method) (torchft.process_group.ProcessGroupWrapper method)

reduce_scatter_tensor_coalesced() (torchft.process_group.ProcessGroup method) (torchft.process_group.ProcessGroupBaby method) (torchft.process_group.ProcessGroupBabyGloo method) (torchft.process_group.ProcessGroupDummy method) (torchft.process_group.ProcessGroupGloo method) (torchft.process_group.ProcessGroupWrapper method) register() (torchft.process_group.ProcessGroup method) register_state_dict_fn() (torchft.manager.Manager method) replica_id (torchft.coordination.QuorumMember attribute) report_error() (torchft.manager.Manager method) (torchft.process_group.ErrorSwallowingProcessGroupWrapper method) report_future_error() (torchft.process_group.FakeProcessGroupWrapper method)

S

send() (torchft.process_group.ProcessGroup method) (torchft.process_group.ProcessGroupBaby method) (torchft.process_group.ProcessGroupDummy method) (torchft.process_group.ProcessGroupWrapper method) send_checkpoint() (torchft.checkpointing.CheckpointTransport method) (torchft.checkpointing.HTTPTransport method) set_state_dict_fns() (torchft.manager.Manager method) set_timeout() (torchft.process_group.ProcessGroup method) (torchft.process_group.ProcessGroupBaby method) (torchft.process_group.ProcessGroupWrapper method) should_commit() (torchft.coordination.ManagerClient method) (torchft.manager.Manager method) shrink_only (torchft.coordination.QuorumMember attribute) shutdown() (torchft.checkpointing.CheckpointTransport method) (torchft.checkpointing.HTTPTransport method) (torchft.coordination.LighthouseServer method) (torchft.coordination.ManagerServer method) (torchft.manager.Manager method) (torchft.process_group.ProcessGroup method) (torchft.process_group.ProcessGroupBaby method) (torchft.process_group.ProcessGroupWrapper method)

size() (torchft.process_group.ManagedProcessGroup method) (torchft.process_group.ProcessGroup method) (torchft.process_group.ProcessGroupBaby method) (torchft.process_group.ProcessGroupDummy method) (torchft.process_group.ProcessGroupWrapper method) start_quorum() (torchft.manager.Manager method) state (torchft.optim.OptimizerWrapper property) state_dict() (torchft.manager.Manager method) (torchft.optim.OptimizerWrapper method) step (torchft.coordination.QuorumMember attribute) step() (torchft.optim.OptimizerWrapper method) store_address (torchft.coordination.QuorumMember attribute) sync() (torchft.local_sgd.LocalSGD method)

T

torchft.checkpointing module torchft.coordination module torchft.data module torchft.ddp module torchft.local_sgd module

torchft.manager module torchft.optim module torchft.parameter_server module torchft.process_group module

U

unregister() (torchft.process_group.ProcessGroup method)

W

wait_quorum() (torchft.manager.Manager method) world_size (torchft.coordination.QuorumMember attribute)

WorldSizeMode (class in torchft.manager) wrap_future() (torchft.manager.Manager method)

Z

zero_grad() (torchft.optim.OptimizerWrapper method)