distributed_strategy.proto

// Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
// Copyright (c) 2021 NVIDIA Corporation. All rights reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//     http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

syntax = "proto2";
package paddle.fleet;

enum Mode {
  COLLECTIVE = 1;
  PS = 2;
  PIPELINE = 3;
  HETER = 4; // support XPU and GPU computing server
}

message RecomputeConfig {
  repeated string checkpoints = 1;
  optional bool enable_offload = 2 [ default = false ];
  repeated int32 checkpoint_shape = 3;
}

message ShardingConfig {
  optional string sharding_segment_strategy = 1
      [ default = 'segment_broadcast_MB' ];
  optional float segment_broadcast_MB = 2 [ default = 32.0 ];
  repeated string segment_anchors = 3;
  optional int32 sharding_degree = 4 [ default = 8 ];
  optional int32 mp_degree = 5 [ default = 1 ];
  optional int32 dp_degree = 6 [ default = 1 ];
  optional bool hybrid_dp = 7 [ default = false ];
  optional int32 gradient_merge_acc_step = 8 [ default = 1 ];
  optional bool optimize_offload = 9 [ default = false ];
  optional bool pp_allreduce_in_optimize = 10 [ default = false ];
  optional int32 pp_degree = 11 [ default = 1 ];
  optional bool optimize_cast = 12 [ default = false ];
  // Optimizer sharding. Temporary plans and may be deprecated
  optional bool _dp_as_optimizer_sharding = 13 [ default = false ];
}

message HybridConfig {
  optional int32 dp_degree = 1 [ default = -1 ];
  optional int32 mp_degree = 2 [ default = 1 ];
  optional int32 pp_degree = 3 [ default = 1 ];
  optional int32 sharding_degree = 4 [ default = 1 ];
}

message AMPConfig {
  optional float init_loss_scaling = 1 [ default = 32768.0 ];
  optional int32 incr_every_n_steps = 2 [ default = 1000 ];
  optional int32 decr_every_n_nan_or_inf = 3 [ default = 2 ];
  optional float incr_ratio = 4 [ default = 2.0 ];
  optional float decr_ratio = 5 [ default = 0.8 ];
  optional bool use_dynamic_loss_scaling = 6 [ default = true ];
  repeated string custom_white_list = 7;
  repeated string custom_black_list = 8;
  repeated string custom_black_varnames = 9;
  optional bool use_pure_fp16 = 10 [ default = false ];
  optional bool use_fp16_guard = 11 [ default = true ];
}

message LocalSGDConfig {
  optional int32 k_steps = 1 [ default = 1 ];
  optional int32 begin_step = 2 [ default = 1 ];
}

message AdaptiveLocalSGDConfig {
  optional int32 init_k_steps = 1 [ default = 1 ];
  optional int32 begin_step = 2 [ default = 1 ];
}

message GradientMergeConfig {
  optional int32 k_steps = 1 [ default = 1 ];
  optional bool avg = 2 [ default = true ];
}

message DGCConfig {
  optional int32 rampup_begin_step = 1 [ default = 0 ];
  optional int32 rampup_step = 2 [ default = 1 ];
  repeated float sparsity = 3;
}

message LarsConfig {
  optional float lars_coeff = 1 [ default = 0.001 ];
  optional float lars_weight_decay = 2 [ default = 0.0005 ];
  optional float epsilon = 3 [ default = 0.0 ];
  repeated string exclude_from_weight_decay = 4;
}

message LambConfig {
  optional float lamb_weight_decay = 1 [ default = 0.01 ];
  repeated string exclude_from_weight_decay = 2;
}

message BuildStrategy {
  optional bool enable_sequential_execution = 1 [ default = false ];
  optional bool fuse_elewise_add_act_ops = 2 [ default = false ];
  optional bool fuse_bn_act_ops = 3 [ default = false ];
  optional bool fuse_relu_depthwise_conv = 4 [ default = false ];
  optional bool fuse_broadcast_ops = 5 [ default = false ];
  optional bool fuse_all_optimizer_ops = 6 [ default = false ];
  optional bool enable_inplace = 7 [ default = false ];
  optional bool enable_backward_optimizer_op_deps = 8 [ default = true ];
  optional bool cache_runtime_context = 9 [ default = false ];
  optional bool fuse_bn_add_act_ops = 10 [ default = true ];
  optional bool enable_auto_fusion = 11 [ default = false ];
  optional bool enable_addto = 12 [ default = false ];
  optional bool fix_op_run_order = 13 [ default = false ];
  optional bool allow_cuda_graph_capture = 14 [ default = false ];
}

message ExecutionStrategy {
  optional int32 num_threads = 1 [ default = 1 ];
  optional int32 num_iteration_per_drop_scope = 2 [ default = 10 ];
  optional int32 num_iteration_per_run = 3 [ default = 1 ];
  optional bool use_thread_barrier = 4 [ default = false ];
}

message GradientScaleConfig {
  // Optional value ['avg', 'sum', 'customized']
  // If avg, loss@grad will be divided by the number of devices,
  // that is, the gradient will be accumulated and averaged among
  // multiple devices.
  // Else if sum, the gradient will accumulated among multiple
  // devices.
  optional string scale_strategy = 1 [ default = 'avg' ];
}

message AsyncConfig {
  optional int32 k_steps = 1 [ default = -1 ];
  optional int32 max_merge_var_num = 2 [ default = 1 ];
  optional int32 send_queue_size = 3 [ default = 16 ];
  optional bool independent_recv_thread = 4 [ default = false ];
  optional int32 min_send_grad_num_before_recv = 5 [ default = 1 ];
  optional int32 thread_pool_size = 6 [ default = 1 ];
  optional int32 send_wait_times = 7 [ default = 1 ];
  optional bool runtime_split_send_recv = 8 [ default = false ];
  optional bool launch_barrier = 9 [ default = true ];
  optional string heter_worker_device_guard = 10 [ default = 'cpu' ];
  optional int32 lr_decay_steps = 11 [ default = 10 ];
  optional int32 use_ps_gpu = 12 [ default = 0 ];
}

message TrainerDescConfig {
  optional string dump_fields_path = 1;
  repeated string dump_fields = 2;
  repeated string dump_param = 3;
  repeated string stat_var_names = 4;
}

message PipelineConfig {
  optional int32 micro_batch_size = 1 [ default = 1 ];
  optional int32 accumulate_steps = 2 [ default = 1 ];
  optional string schedule_mode = 3 [ default = '1F1B' ];
  optional bool p2p_cache_shape = 4 [ default = true ];
}

message TensorParallelConfig {
  optional int32 tensor_parallel_degree = 1 [ default = 1 ];
  optional int32 tensor_init_seed = 2 [ default = -1 ];
}

message DistributedStrategy {
  // bool options
  optional Mode mode = 1 [ default = COLLECTIVE ];
  optional bool amp = 2 [ default = false ];
  optional bool recompute = 3 [ default = false ];
  optional bool localsgd = 4 [ default = false ];
  optional bool dgc = 5 [ default = false ];
  optional bool gradient_merge = 6 [ default = false ];
  optional bool lars = 7 [ default = false ];
  optional bool lamb = 8 [ default = false ];
  optional bool pipeline = 9 [ default = false ];
  optional bool elastic = 10 [ default = false ];
  optional bool auto = 11 [ default = false ];
  optional bool a_sync = 12 [ default = true ];
  optional bool sync_nccl_allreduce = 13 [ default = true ];
  optional int32 nccl_comm_num = 14 [ default = 1 ];
  optional bool use_hierarchical_allreduce = 15 [ default = false ];
  optional int32 hierarchical_allreduce_inter_nranks = 16 [ default = 1 ];
  optional bool sync_batch_norm = 17 [ default = false ];
  optional bool fuse_all_reduce_ops = 18 [ default = true ];
  optional int32 fuse_grad_size_in_MB = 19 [ default = 32 ];
  optional float fuse_grad_size_in_TFLOPS = 20 [ default = 50 ];
  optional bool cudnn_exhaustive_search = 21 [ default = false ];
  optional int32 conv_workspace_size_limit = 22 [ default = 512 ];
  optional bool cudnn_batchnorm_spatial_persistent = 23 [ default = false ];
  optional bool adaptive_localsgd = 24 [ default = false ];
  optional bool fp16_allreduce = 25 [ default = false ];
  optional bool sharding = 26 [ default = false ];
  optional float last_comm_group_size_MB = 27 [ default = 1 ];
  optional bool find_unused_parameters = 28 [ default = false ];
  optional bool tensor_parallel = 29 [ default = false ];
  optional bool without_graph_optimization = 30 [ default = false ];
  optional int32 fuse_grad_size_in_num = 31 [ default = 8 ];
  optional bool calc_comm_same_stream = 32 [ default = false ];
  optional bool asp = 33 [ default = false ];
  optional bool fuse_grad_merge = 34 [ default = false ];
  optional bool semi_auto = 35 [ default = false ];

  optional RecomputeConfig recompute_configs = 101;
  optional AMPConfig amp_configs = 102;
  optional LocalSGDConfig localsgd_configs = 103;
  optional GradientMergeConfig gradient_merge_configs = 104;
  optional DGCConfig dgc_configs = 105;
  optional PipelineConfig pipeline_configs = 106;
  optional AsyncConfig a_sync_configs = 107;
  optional LarsConfig lars_configs = 108;
  optional LambConfig lamb_configs = 109;
  optional AdaptiveLocalSGDConfig adaptive_localsgd_configs = 110;
  optional ShardingConfig sharding_configs = 111;
  optional HybridConfig hybrid_configs = 112;
  optional TensorParallelConfig tensor_parallel_configs = 113;
  optional TrainerDescConfig trainer_desc_configs = 114;
  optional BuildStrategy build_strategy = 201;
  optional ExecutionStrategy execution_strategy = 202;
  optional GradientScaleConfig gradient_scale_configs = 203;
}

message DistributedJobInfo {
  optional int32 worker_num = 1;
  optional int32 server_num = 2;
  repeated string worker_ips = 3;
  repeated string server_endpoints = 4;
  optional string origin_startup = 5;
  optional string origin_main = 6; // without backpropagation and optimization
  optional string distributed_main = 7; // with backpropagation and optimization
  optional string optimizer_name = 8;   // optimizer name
  optional DistributedStrategy strategy = 101;
}