distributed_strategy.proto 6.2 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
// Copyright (c) 2020 PaddlePaddle Authors. 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
}

25 26 27 28 29 30 31 32 33
message RecomputeConfig { repeated string checkpoints = 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 ];
34 35 36
  repeated string custom_white_list = 7;
  repeated string custom_black_list = 8;
  repeated string custom_black_varnames = 9;
37
}
38

39 40 41 42
message LocalSGDConfig {
  optional int32 k_steps = 1 [ default = 1 ];
  optional int32 begin_step = 2 [ default = 1 ];
}
43

44 45 46 47 48
message AdaptiveLocalSGDConfig {
  optional int32 init_k_steps = 1 [ default = 1 ];
  optional int32 begin_step = 2 [ default = 1 ];
}

49 50 51 52 53
message GradientMergeConfig {
  optional int32 k_steps = 1 [ default = 1 ];
  optional bool avg = 2 [ default = true ];
}

54 55 56 57 58 59
message DGCConfig {
  optional int32 rampup_begin_step = 1 [ default = 0 ];
  optional int32 rampup_step = 2 [ default = 1 ];
  repeated float sparsity = 3;
}

60 61 62
message LarsConfig {
  optional float lars_coeff = 1 [ default = 0.001 ];
  optional float lars_weight_decay = 2 [ default = 0.0005 ];
63 64
  optional float epsilon = 3 [ default = 0.0 ];
  repeated string exclude_from_weight_decay = 4;
65 66 67
}

message LambConfig {
68 69
  optional float lamb_weight_decay = 1 [ default = 0.01 ];
  repeated string exclude_from_weight_decay = 2;
70 71
}

72 73 74 75 76 77 78 79 80 81 82
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 ];
}
83

84 85 86 87 88 89 90 91
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 AsyncConfig {
92
  optional int32 k_steps = 1 [ default = -1 ];
93 94 95 96 97 98 99
  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 ];
C
Chengmo 已提交
100
  optional bool launch_barrier = 9 [ default = true ];
101
  optional string heter_worker_device_guard = 10 [ default = 'cpu' ];
102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118
}

message PipelineConfig { optional int32 micro_batch = 1 [ 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 ];
D
Dong Daxiang 已提交
119
  optional bool a_sync = 12 [ default = true ];
120 121 122 123 124 125
  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 ];
126 127
  optional int32 fuse_grad_size_in_MB = 19 [ default = 32 ];
  optional float fuse_grad_size_in_TFLOPS = 20 [ default = 50 ];
128 129 130
  optional bool cudnn_exhaustive_search = 21 [ default = true ];
  optional int32 conv_workspace_size_limit = 22 [ default = 4000 ];
  optional bool cudnn_batchnorm_spatial_persistent = 23 [ default = true ];
131
  optional bool adaptive_localsgd = 24 [ default = false ];
132
  optional bool fp16_allreduce = 25 [ default = false ];
133

134 135 136 137
  optional RecomputeConfig recompute_configs = 101;
  optional AMPConfig amp_configs = 102;
  optional LocalSGDConfig localsgd_configs = 103;
  optional GradientMergeConfig gradient_merge_configs = 104;
138
  optional DGCConfig dgc_configs = 105;
139
  optional PipelineConfig pipeline_configs = 106;
D
Dong Daxiang 已提交
140
  optional AsyncConfig a_sync_configs = 107;
141 142
  optional LarsConfig lars_configs = 108;
  optional LambConfig lamb_configs = 109;
143
  optional AdaptiveLocalSGDConfig adaptive_localsgd_configs = 110;
144 145
  optional BuildStrategy build_strategy = 201;
  optional ExecutionStrategy execution_strategy = 202;
146 147 148 149 150 151 152 153 154 155 156 157 158
}

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;
}