# 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. import collections import logging import math import warnings from functools import reduce import paddle from paddle.framework import core from paddle.incubate.distributed.fleet.parameter_server.ir import vars_metatools from paddle.incubate.distributed.fleet.parameter_server.ir.ps_dispatcher import ( RoundRobin, ) from paddle.incubate.distributed.fleet.parameter_server.mode import ( DistributedMode, ) OP_NAME_SCOPE = "op_namescope" CLIP_OP_NAME_SCOPE = "gradient_clip" STEP_COUNTER = "@PS_STEP_COUNTER@" LEARNING_RATE_DECAY_COUNTER = "@LR_DECAY_COUNTER@" OP_ROLE_VAR_ATTR_NAME = core.op_proto_and_checker_maker.kOpRoleVarAttrName() RPC_OP_ROLE_ATTR_NAME = core.op_proto_and_checker_maker.kOpRoleAttrName() RPC_OP_ROLE_ATTR_VALUE = core.op_proto_and_checker_maker.OpRole.RPC op_role_attr_name = core.op_proto_and_checker_maker.kOpRoleAttrName() LR_SCHED_OP_ROLE_ATTR_VALUE = core.op_proto_and_checker_maker.OpRole.LRSched OPT_OP_ROLE_ATTR_VALUE = core.op_proto_and_checker_maker.OpRole.Optimize SPARSE_OP_LIST = ["lookup_table", "lookup_table_v2"] SPARSE_OP_TYPE_DICT = {"lookup_table": "W", "lookup_table_v2": "W"} def _get_lr_ops(program): lr_ops = [] for index, op in enumerate(program.global_block().ops): role_id = int(op.attr(RPC_OP_ROLE_ATTR_NAME)) if role_id == int(LR_SCHED_OP_ROLE_ATTR_VALUE) or role_id == int( LR_SCHED_OP_ROLE_ATTR_VALUE ) | int(OPT_OP_ROLE_ATTR_VALUE): lr_ops.append(op) return lr_ops def _has_global_step(lr_ops): if len(lr_ops) > 0: for idx, op in enumerate(lr_ops): if op.type != 'increment': continue counter = op.input("X")[0] if counter == LEARNING_RATE_DECAY_COUNTER: return True return False def is_sparse_op(op): if ( op.type in SPARSE_OP_LIST and op.attr('is_sparse') is True and op.attr('is_distributed') is False ): return True if ( op.type == "distributed_lookup_table" and op.attr('is_distributed') is False ): return True return False def is_distributed_sparse_op(op): if op.type in SPARSE_OP_LIST and op.attr('is_distributed') is True: return True if ( op.type == "distributed_lookup_table" and op.attr('is_distributed') is True ): return True return False def get_sparse_tablename(op): return op.input("W")[0] def get_sparse_tablenames(program, is_distributed): tablenames = set() if is_distributed: for op in program.global_block().ops: if is_distributed_sparse_op(op): tablenames.add(get_sparse_tablename(op)) else: for op in program.global_block().ops: if is_sparse_op(op): tablenames.add(get_sparse_tablename(op)) return list(tablenames) class MergedVariable: def __init__(self, merged, ordered, offsets): self.merged_var = merged self.ordered_vars = ordered self.offsets = offsets def Singleton(cls): _instance = {} def _singleton(*args, **kargs): if cls not in _instance: _instance[cls] = cls(*args, **kargs) return _instance[cls] return _singleton @Singleton class CompileTimeStrategy: def __init__(self, main_program, startup_program, strategy, role_maker): self.min_block_size = 81920 self.origin_main_program = main_program self.origin_startup_program = startup_program self.origin_ps_main_program = main_program self.origin_ps_startup_program = startup_program self.strategy = strategy self.role_maker = role_maker self.use_ps_gpu = False try: self.is_heter_ps_mode = role_maker._is_heter_parameter_server_mode except: warnings.warn( "Using paddle.distributed.fleet instead of paddle.fluid.incubate.fleet" ) self.is_heter_ps_mode = False self.origin_sparse_pairs = [] self.origin_dense_pairs = [] self.merged_variables_pairs = [] self.merged_dense_pairs = [] self.merged_sparse_pairs = [] self.merged_variable_map = {} self.param_name_to_grad_name = {} self.grad_name_to_param_name = {} self.param_grad_ep_mapping = collections.OrderedDict() self.grad_param_mapping = collections.OrderedDict() self._build_var_distributed() self.tensor_table_dict = {} # for heter-ps save variables self.origin_merged_variables_pairs = list(self.merged_variables_pairs) self.origin_merged_dense_pairs = list(self.merged_dense_pairs) self.origin_merged_sparse_pairs = list(self.merged_sparse_pairs) def get_distributed_mode(self): trainer = self.strategy.get_trainer_runtime_config() return trainer.mode def is_sync_mode(self): trainer = self.strategy.get_trainer_runtime_config() return trainer.mode == DistributedMode.SYNC def is_geo_mode(self): trainer = self.strategy.get_trainer_runtime_config() return trainer.mode == DistributedMode.GEO def is_async_mode(self): trainer = self.strategy.get_trainer_runtime_config() return trainer.mode == DistributedMode.ASYNC def get_role_id(self): try: return self.role_maker._role_id() except Exception: return self.role_maker.role_id() def get_trainers(self): try: return self.role_maker._worker_num() except Exception: return self.role_maker.worker_num() def get_ps_endpoint(self): try: return self.role_maker._get_pserver_endpoints()[self.get_role_id()] except Exception: return self.role_maker.get_pserver_endpoints()[self.get_role_id()] def get_ps_endpoints(self): try: return self.role_maker._get_pserver_endpoints() except Exception: return self.role_maker.get_pserver_endpoints() def get_heter_worker_endpoints(self): try: return self.role_maker._get_heter_worker_endpoints() except Exception: return self.role_maker.get_heter_worker_endpoints() def get_next_stage_trainers(self): try: return self.role_maker._get_next_trainers() except Exception: return self.role_maker.get_next_trainers() def get_heter_worker_endpoint(self): try: return self.role_maker._get_heter_worker_endpoint() except Exception: return self.role_maker.get_heter_worker_endpoint() def get_trainer_endpoints(self): try: return self.role_maker._get_trainer_endpoints() except Exception: return self.role_maker.get_trainer_endpoints() def get_trainer_endpoint(self): try: return self.role_maker._get_trainer_endpoint() except Exception: return self.role_maker.get_trainer_endpoint() def get_previous_stage_trainers(self): try: return self.role_maker._get_previous_trainers() except Exception: return self.role_maker.get_previous_trainers() def get_origin_programs(self): return self.origin_main_program, self.origin_startup_program def get_origin_main_program(self): return self.origin_main_program def get_origin_startup_program(self): return self.origin_startup_program def set_origin_ps_main_program(self, program): self.origin_ps_main_program = program def set_origin_ps_startup_program(self, program): self.origin_ps_startup_program = program def get_origin_ps_main_program(self): return self.origin_ps_main_program def get_origin_ps_startup_program(self): return self.origin_ps_startup_program def add_tensor_table( self, feed_var_name, fetch_var_name="", startup_program=None, main_program=None, tensor_table_class="", ): self.tensor_table_dict[feed_var_name] = {} self.tensor_table_dict[feed_var_name]["feed_var_name"] = feed_var_name self.tensor_table_dict[feed_var_name]["fetch_var_name"] = fetch_var_name self.tensor_table_dict[feed_var_name][ "startup_program" ] = startup_program self.tensor_table_dict[feed_var_name]["main_program"] = main_program self.tensor_table_dict[feed_var_name][ "tensor_table_class" ] = tensor_table_class def get_tensor_table_dict(self): return self.tensor_table_dict def get_sparse_varname_on_ps(self, is_distributed, endpoint=None): if not endpoint: endpoint = self.get_ps_endpoint() varnames = get_sparse_tablenames( self.get_origin_main_program(), is_distributed ) ps_sparse_varnames = [] for varname in varnames: tables = self.get_var_distributed(varname, True) for i in range(len(tables)): table, ep, _ = tables[i] if ep == endpoint: ps_sparse_varnames.append(table) return ps_sparse_varnames def get_optimize_varname_on_ps(self, param_name): origin_param_name, _, _ = _get_varname_parts(param_name) optimize_var_names = [] for op in self.get_origin_main_program().global_block().ops: # check all optimizer op if int(op.all_attrs()["op_role"]) == 2: # check param name if op.input("Param")[0] != origin_param_name: continue # check all input for key in op.input_names: if key in [ "Param", "Grad", "LearningRate", "Beta1Tensor", "Beta2Tensor", ]: continue # check varibale shape related param, e.g: Moment1 optimize_var_names += ( self._get_optimizer_param_related_var_name( op, op.type, key ) ) return optimize_var_names def _get_optimizer_param_related_var_name(self, op, op_type, varkey): """ Returns the names for optimizer inputs that need to be load """ related_var_names = [] if op_type == "adam": if varkey in ["Moment1", "Moment2"]: related_var_names.append(op.input(varkey)[0]) elif op_type == "adagrad": if varkey == "Moment": related_var_names.append(op.input(varkey)[0]) elif op_type in ["momentum", "lars_momentum"]: if varkey == "Velocity": related_var_names.append(op.input(varkey)[0]) elif op_type == "rmsprop": if varkey in ["Moment", "MeanSquare"]: related_var_names.append(op.input(varkey)[0]) elif op_type == "ftrl": if varkey in ["SquaredAccumulator", "LinearAccumulator"]: related_var_names.append(op.input(varkey)[0]) elif op_type == "sgd": pass else: raise ValueError( "Not supported optimizer for distributed training: %s" % op_type ) return related_var_names def build_ctx( self, vars, mapping, is_grad, is_sparse, is_send, is_distributed=False ): def get_grad_var_ep(slices): names = [] eps = [] sections = [] for slice in slices: if self.is_geo_mode(): if is_send: names.append("{}.delta".format(slice.name)) else: names.append(slice.name) elif ( is_grad and self.is_sync_mode() and self.get_trainers() > 1 ): names.append( "{}.trainer_{}".format(slice.name, self.get_role_id()) ) else: names.append(slice.name) sections.append(slice.shape[0]) for ep, pairs in self.param_grad_ep_mapping.items(): params, grads = pairs["params"], pairs["grads"] for var in params + grads: if slice.name == var.name: eps.append(ep) break return names, eps, sections if isinstance(vars, MergedVariable): name = vars.merged_var.name slices = mapping[name] names, eps, sections = get_grad_var_ep(slices) origin_varnames = [var.name for var in vars.ordered_vars] else: name = vars.name slices = mapping[name] names, eps, sections = get_grad_var_ep(slices) origin_varnames = [vars.name] trainer_id = self.get_role_id() aggregate = True ctx = core.CommContext( name, names, eps, sections, origin_varnames, trainer_id, aggregate, is_sparse, is_distributed, [], ) return ctx def get_trainer_send_context(self): send_ctx = {} distibuted_varnames = get_sparse_tablenames( self.origin_main_program, True ) idx = 0 if not self.is_geo_mode(): for merged in self.merged_dense_pairs: grad = merged[1] ctx = self.build_ctx( grad, self.grad_var_mapping, True, False, True ) send_ctx[ctx.var_name()] = ctx for merged in self.merged_sparse_pairs: param = merged[0] grad = merged[1] param_name = param.merged_var.name is_distributed = ( True if param_name in distibuted_varnames else False ) ctx = self.build_ctx( grad, self.grad_var_mapping, True, True, True, is_distributed, ) send_ctx[ctx.var_name()] = ctx idx += 1 if self.is_async_mode(): name, ctx = self._step_ctx(idx) send_ctx[name] = ctx else: for pairs in self.origin_sparse_pairs: param, grad = pairs param_name = param.name is_distributed = ( True if param_name in distibuted_varnames else False ) param_ctx = self.build_ctx( param, self.param_var_mapping, False, True, True, is_distributed, ) grad_ctx = self.build_ctx( grad, self.grad_var_mapping, True, True, True, is_distributed, ) ctx = core.CommContext( param_ctx.var_name(), param_ctx.split_varnames(), param_ctx.split_endpoints(), param_ctx.sections(), grad_ctx.origin_varnames(), param_ctx.trainer_id(), param_ctx.aggregate(), param_ctx.is_sparse(), param_ctx.is_distributed(), [], ) send_ctx[ctx.var_name()] = ctx idx += 1 name, ctx = self._step_ctx(idx) send_ctx[name] = ctx return send_ctx def get_communicator_send_context(self): send_ctx = {} distibuted_varnames = get_sparse_tablenames( self.origin_main_program, True ) idx = 0 if self.is_geo_mode(): for pairs in self.merged_dense_pairs: param = pairs[0] ctx = self.build_ctx( param, self.param_var_mapping, False, False, True ) send_ctx[ctx.var_name()] = ctx for pairs in self.merged_sparse_pairs: param = pairs[0] param_name = param.merged_var.name is_distributed = ( True if param_name in distibuted_varnames else False ) ctx = self.build_ctx( param, self.param_var_mapping, False, True, True, is_distributed, ) send_ctx[ctx.var_name()] = ctx idx += 1 name, ctx = self._step_ctx(idx) send_ctx[name] = ctx else: for merged in self.merged_dense_pairs: grad = merged[1] ctx = self.build_ctx( grad, self.grad_var_mapping, True, False, True ) send_ctx[ctx.var_name()] = ctx for merged in self.merged_sparse_pairs: param, grad = merged param_name = param.merged_var.name is_distributed = ( True if param_name in distibuted_varnames else False ) ctx = self.build_ctx( grad, self.grad_var_mapping, True, True, True, is_distributed, ) send_ctx[ctx.var_name()] = ctx idx += 1 name, ctx = self._step_ctx(idx) send_ctx[name] = ctx return send_ctx def get_communicator_recv_context( self, recv_type=1, use_origin_program=False ): # recv_type # 1 : DENSE 2. SPARSE 3. DISTRIBUTED 4. ALL distibuted_varnames = get_sparse_tablenames( self.origin_main_program, True ) sparse_varnames = [] for pairs in self.origin_sparse_pairs: param, grad = pairs sparse_varnames.append(param.name) dense_recv_ctx = {} sparse_recv_ctx = {} distributed_recv_ctx = {} variables_pairs = ( self.merged_variables_pairs if not use_origin_program else self.origin_merged_variables_pairs ) for merged in variables_pairs: params = merged[0] if params.merged_var.name in sparse_varnames: continue ctx = self.build_ctx( params, self.param_var_mapping, False, False, False, False ) dense_recv_ctx[ctx.var_name()] = ctx for pairs in self.origin_sparse_pairs: param, grad = pairs if param.name in distibuted_varnames: ctx = self.build_ctx( param, self.param_var_mapping, False, True, False, True ) distributed_recv_ctx[ctx.var_name()] = ctx else: ctx = self.build_ctx( param, self.param_var_mapping, False, True, False, False ) sparse_recv_ctx[ctx.var_name()] = ctx if recv_type == 1: return dense_recv_ctx if recv_type == 2: return sparse_recv_ctx if recv_type == 3: return distributed_recv_ctx if recv_type == 4: dense_recv_ctx.update(sparse_recv_ctx) dense_recv_ctx.update(distributed_recv_ctx) return dense_recv_ctx assert ValueError( "recv_type can only be 1/2/3/4, 1 : DENSE 2. SPARSE 3. DISTRIBUTED 4. ALL" ) def get_the_one_trainer_send_context(self, split_dense_table): if self.is_geo_mode(): send_ctx = {} trainer_id = self.get_role_id() idx = 0 distibuted_varnames = get_sparse_tablenames( self.origin_main_program, True ) for merged in self.merged_sparse_pairs: param, grad = merged grad_name = grad.merged_var.name param_name = param.merged_var.name is_distributed = ( True if param_name in distibuted_varnames else False ) var = self.origin_main_program.global_block().vars[ grad.merged_var.name ] var_numel = reduce(lambda x, y: x * y, var.shape[1:]) sparse_ctx = core.CommContext( grad_name, [grad_name], ["127.0.0.1:6071"], [var_numel], [grad_name], trainer_id, True, True, is_distributed, idx, False, False, -1, [], ) idx += 1 send_ctx[sparse_ctx.var_name()] = sparse_ctx if len(send_ctx) == 0: raise ValueError( "GeoSGD require sparse parameters in your net." ) if len(self.tensor_table_dict) > 0 and self.role_maker._is_worker(): name, ctx = self._step_ctx(idx) send_ctx[name] = ctx return send_ctx else: return self.get_the_one_send_context(split_dense_table) def get_dense_send_context( self, send_ctx, idx, merged_dense_pairs, trainer_id, split_dense_table=False, ): if len(merged_dense_pairs) < 1: return idx if not split_dense_table: origin_varnames = [] var_numel = 0 for merged in merged_dense_pairs: grad = merged[1] origin_varnames.append(grad.merged_var.name) var = self.origin_main_program.global_block().vars[ grad.merged_var.name ] var_numel += reduce(lambda x, y: x * y, var.shape) grad_name = "Dense@Grad" trainer_id = self.get_role_id() aggregate = True dense_ctx = core.CommContext( grad_name, [grad_name], ["127.0.0.1:6071"], [var_numel], origin_varnames, trainer_id, aggregate, False, False, idx, False, False, -1, [], ) send_ctx[grad_name] = dense_ctx idx += 1 else: for merged in merged_dense_pairs: grad = merged[1] origin_varname = grad.merged_var.name var = self.origin_main_program.global_block().vars[ origin_varname ] var_numel = reduce(lambda x, y: x * y, var.shape) grad_name = origin_varname aggregate = True dense_ctx = core.CommContext( grad_name, [grad_name], ["127.0.0.1:6071"], [var_numel], [origin_varname], trainer_id, aggregate, False, False, idx, False, False, -1, [], ) send_ctx[grad_name] = dense_ctx idx += 1 return idx def get_the_one_send_context( self, split_dense_table=False, use_origin_program=False, ep_list=None ): if ep_list is None: ep_list = ["127.0.0.1:6071"] send_ctx = {} trainer_id = self.get_role_id() idx = 0 merged_dense_pairs = ( self.origin_merged_dense_pairs if use_origin_program else self.merged_dense_pairs ) merged_sparse_pairs = ( self.origin_merged_sparse_pairs if use_origin_program else self.merged_sparse_pairs ) idx += self.get_dense_send_context( send_ctx, idx, merged_dense_pairs, trainer_id, split_dense_table ) distibuted_varnames = get_sparse_tablenames( self.origin_main_program, True ) for merged in merged_sparse_pairs: param, grad = merged grad_name = grad.merged_var.name param_name = param.merged_var.name splited_varname = [] for i in range(len(ep_list)): splited_varname.append("{}.block{}".format(param_name, i)) is_distributed = ( True if param_name in distibuted_varnames else False ) var = self.origin_main_program.global_block().vars[ grad.merged_var.name ] shape = list(var.shape) shape[0] = 0 if is_distributed else shape[0] sparse_ctx = core.CommContext( grad_name, splited_varname, ep_list, shape, [grad_name], trainer_id, True, True, is_distributed, idx, False, False, -1, [], ) idx += 1 send_ctx[sparse_ctx.var_name()] = sparse_ctx if len(self.tensor_table_dict) > 0 and self.role_maker._is_worker(): name, ctx = self._step_ctx(idx) send_ctx[name] = ctx return send_ctx def get_the_one_recv_context( self, is_dense=True, split_dense_table=False, use_origin_program=False ): recv_id_maps = {} if is_dense: send_ctx = self.get_the_one_send_context( split_dense_table=split_dense_table, use_origin_program=use_origin_program, ) for idx, (name, ctx) in enumerate(send_ctx.items()): if ctx.is_sparse(): continue if ctx.is_tensor_table(): continue origin_grad_varnames = ctx.origin_varnames() param_names = [] for grad_varname in origin_grad_varnames: param_name = self.grad_name_to_param_name[grad_varname] param_names.append(param_name) recv_id_maps[ctx.table_id()] = param_names else: send_ctx = self.get_the_one_send_context() for idx, (name, ctx) in enumerate(send_ctx.items()): if not ctx.is_sparse(): continue origin_grad_varnames = ctx.origin_varnames() param_names = [] for grad_varname in origin_grad_varnames: param_name = self.grad_name_to_param_name[grad_varname] param_names.append(param_name) recv_id_maps[ctx.table_id()] = param_names return recv_id_maps def get_server_runtime_config(self): return self.strategy.get_server_runtime_config() def get_var_distributed(self, varname, is_param): var_distributed = [] offset = 0 if is_param: params = self.param_var_mapping[varname] param_varnames = [var.name for var in params] for ep, pairs in self.param_grad_ep_mapping.items(): for p in pairs["params"]: if p.name in param_varnames: offset += p.shape[0] var_distributed.append((p.name, ep, p.shape[0])) else: grads = self.grad_var_mapping[varname] grad_varnames = [var.name for var in grads] for ep, pairs in self.param_grad_ep_mapping.items(): for g in pairs["grads"]: if g.name in grad_varnames: var_distributed.append((g.name, ep, g.shape[0])) return var_distributed def _step_ctx(self, idx): name = STEP_COUNTER trainer_id = self.get_role_id() endpoints = self.get_ps_endpoints() sections = [1] * len(endpoints) names = [name] * len(endpoints) ctx = core.CommContext( name, names, endpoints, sections, [name], trainer_id, True, False, False, idx, True, False, -1, [], ) return name, ctx def _create_vars_from_blocklist(self, block_list): """ Create vars for each split. NOTE: only grads need to be named for different trainers, use add_trainer_suffix to rename the grad vars. Args: block_list (list[(varname, block_id, block_size)]): List of gradient blocks. add_trainer_suffix (Bool): Add trainer suffix to new variable's name if set True. Returns: var_mapping (collections.OrderedDict(varname->[new_varname_variable])):A dict mapping from original var name to each var split. """ # varname->[(block_id, current_block_size)] block_map = collections.OrderedDict() var_mapping = collections.OrderedDict() for block_str in block_list: varname, offset, size = block_str.split(":") if varname not in block_map: block_map[varname] = [] block_map[varname].append((int(offset), int(size))) for varname, split in block_map.items(): orig_var = self.merged_variable_map[varname] if len(split) == 1: var_mapping[varname] = [orig_var] self.var_distributed.add_distributed_var( origin_var=orig_var, slice_var=orig_var, block_id=0, offset=0, is_slice=False, vtype="Param", ) else: var_mapping[varname] = [] orig_shape = orig_var.shape orig_dim1_flatten = 1 if len(orig_shape) >= 2: orig_dim1_flatten = reduce( lambda x, y: x * y, orig_shape[1:] ) for i, block in enumerate(split): size = block[1] rows = size // orig_dim1_flatten splited_shape = [rows] if len(orig_shape) >= 2: splited_shape.extend(orig_shape[1:]) new_var_name = "%s.block%d" % (varname, i) slice_var = vars_metatools.VarStruct( name=new_var_name, shape=splited_shape, dtype=orig_var.dtype, type=orig_var.type, lod_level=orig_var.lod_level, persistable=False, ) var_mapping[varname].append(slice_var) self.var_distributed.add_distributed_var( origin_var=orig_var, slice_var=slice_var, block_id=i, offset=-1, is_slice=False, vtype="Param", ) return var_mapping def _dispatcher(self): ps_dispatcher = RoundRobin(self.get_ps_endpoints()) ps_dispatcher.reset() grad_var_mapping_items = list(self.grad_var_mapping.items()) sparse_gradnames = [grad.name for _, grad in self.origin_sparse_pairs] for grad_varname, splited_vars in grad_var_mapping_items: if grad_varname in sparse_gradnames: continue send_vars = [] for _, var in enumerate(splited_vars): send_vars.append(var) recv_vars = [] for _, var in enumerate(send_vars): recv_vars.append(self.grad_param_mapping[var]) eps = ps_dispatcher.dispatch(recv_vars) for i, ep in enumerate(eps): self.param_grad_ep_mapping[ep]["params"].append(recv_vars[i]) self.param_grad_ep_mapping[ep]["grads"].append(send_vars[i]) for grad_varname, splited_vars in grad_var_mapping_items: if grad_varname not in sparse_gradnames: continue ps_dispatcher.reset() send_vars = [] for _, var in enumerate(splited_vars): send_vars.append(var) recv_vars = [] for _, var in enumerate(send_vars): recv_vars.append(self.grad_param_mapping[var]) eps = ps_dispatcher.dispatch(recv_vars) for i, ep in enumerate(eps): self.param_grad_ep_mapping[ep]["params"].append(recv_vars[i]) self.param_grad_ep_mapping[ep]["grads"].append(send_vars[i]) def _slice_variable( self, var_list, slice_count, min_block_size, uniform=False ): """ We may need to split dense tensor to one or more blocks and put them equally onto parameter server. One block is a sub-tensor aligned by dim[0] of the tensor. We need to have a minimal block size so that the calculations in the parameter server side can gain better performance. By default minimum block size 8K elements (maybe 16bit or 32bit or 64bit). Args: var_list (list): List of variables. slice_count (int): Numel of count that variables will be sliced, which could be the pserver services' count. min_block_size (int): Minimum split block size. Returns: blocks (list[(varname, block_id, current_block_size)]): A list of VarBlocks. Each VarBlock specifies a shard of the var. """ blocks = [] for var in var_list: if not uniform: var_numel = reduce(lambda x, y: x * y, var.shape) split_count = 1 if min_block_size == -1: split_count = 1 else: split_count = slice_count max_pserver_count = int( math.floor(var_numel / float(min_block_size)) ) if max_pserver_count == 0: max_pserver_count = 1 if max_pserver_count < slice_count: split_count = max_pserver_count block_size = int(math.ceil(var_numel / float(split_count))) if len(var.shape) >= 2: # align by dim1(width) dim1 = reduce(lambda x, y: x * y, var.shape[1:]) remains = block_size % dim1 if remains != 0: block_size += dim1 - remains # update split_count after aligning split_count = int(math.ceil(var_numel / float(block_size))) for block_id in range(split_count): curr_block_size = min( block_size, var_numel - ((block_id) * block_size) ) block = vars_metatools.VarBlock( var.name, block_id, curr_block_size ) blocks.append(str(block)) else: block_size = var.shape[0] / slice_count remainder = var.shape[0] % slice_count if block_size == 0: dim0s = [block_size] * remainder else: dim0s = [block_size] * slice_count for i in range(remainder): dim0s[i] = dim0s[i] + 1 dim1 = reduce(lambda x, y: x * y, var.shape[1:]) for block_id in range(len(dim0s)): numel = dim0s[block_id] * dim1 block = vars_metatools.VarBlock(var.name, block_id, numel) blocks.append(str(block)) return blocks def _get_param_grad_blocks(self, pairs, min_block_size, uniform=False): param_list = [] grad_list = [] param_grad_set = set() for p, g in pairs: # todo(tangwei12) skip parameter marked not trainable # if type(p) == Parameter and p.trainable == False: # continue p = p.merged_var g = g.merged_var if p.name not in param_grad_set: param_list.append(p) param_grad_set.add(p.name) if g.name not in param_grad_set: grad_list.append(g) param_grad_set.add(g.name) # when we slice var up into blocks, we will slice the var according to # pserver services' count. A pserver may have two or more listening ports. grad_blocks = self._slice_variable( grad_list, len(self.get_ps_endpoints()), min_block_size, uniform ) param_blocks = self._slice_variable( param_list, len(self.get_ps_endpoints()), min_block_size, uniform ) return param_blocks, grad_blocks def _var_slice_and_distribute(self): # update these mappings for further transpile: # 1. param_var_mapping : param var name->[split params vars] # 2. grad_var_mapping : grad var name->[split grads vars] # 3. grad_param_mapping : grad.blockx->param.blockx # 4. param_grad_ep_mapping : ep->{"params" : [], "grads" : [] } dps, dgs = self._get_param_grad_blocks( self.merged_dense_pairs, self.min_block_size, False ) sps, sgs = self._get_param_grad_blocks( self.merged_sparse_pairs, self.min_block_size, True ) param_blocks = dps + sps grad_blocks = dgs + sgs assert len(grad_blocks) == len(param_blocks) # origin_param_name->[splited_param_vars] self.param_var_mapping = self._create_vars_from_blocklist(param_blocks) self.grad_var_mapping = self._create_vars_from_blocklist(grad_blocks) # dict(grad_splited_var->param_splited_var) self.grad_param_mapping = collections.OrderedDict() for g, p in zip(grad_blocks, param_blocks): g_name, g_bid, _ = g.split(":") p_name, p_bid, _ = p.split(":") self.grad_param_mapping[ self.grad_var_mapping[g_name][int(g_bid)] ] = self.param_var_mapping[p_name][int(p_bid)] print_maps = {} for k, v in self.grad_param_mapping.items(): print_maps[str(k)] = str(v) # create mapping of endpoint->split var to create pserver side program self.param_grad_ep_mapping = collections.OrderedDict() [ self.param_grad_ep_mapping.update({ep: {"params": [], "grads": []}}) for ep in self.get_ps_endpoints() ] def _build_var_distributed(self): self.var_distributed = vars_metatools.VarsDistributed() sparse_pairs, dense_pairs = self.get_param_grads() origin_for_sparse = [] origin_for_dense = [] param_name_grad_name = dict() grad_name_to_param_name = dict() for param, grad in sparse_pairs: param = vars_metatools.create_var_struct(param) grad = vars_metatools.create_var_struct(grad) origin_for_sparse.append((param, grad)) for param, grad in dense_pairs: param = vars_metatools.create_var_struct(param) grad = vars_metatools.create_var_struct(grad) origin_for_dense.append((param, grad)) for dense_pair in origin_for_dense: param, grad = dense_pair m_param = MergedVariable(param, [param], [0]) m_grad = MergedVariable(grad, [grad], [0]) self.merged_variables_pairs.append((m_param, m_grad)) self.merged_dense_pairs.append((m_param, m_grad)) for sparse_pair in origin_for_sparse: param, grad = sparse_pair m_param = MergedVariable(param, [param], [0]) m_grad = MergedVariable(grad, [grad], [0]) self.merged_variables_pairs.append((m_param, m_grad)) self.merged_sparse_pairs.append((m_param, m_grad)) for merged in self.merged_variables_pairs: m_param, m_grad = merged self.merged_variable_map[ m_param.merged_var.name ] = m_param.merged_var self.merged_variable_map[m_grad.merged_var.name] = m_grad.merged_var param_merges = [] param_merges.extend(origin_for_sparse) param_merges.extend(origin_for_dense) for param, grad in param_merges: param_name_grad_name[param.name] = grad.name grad_name_to_param_name[grad.name] = param.name self.origin_sparse_pairs = origin_for_sparse self.origin_dense_pairs = origin_for_dense self.param_name_to_grad_name = param_name_grad_name self.grad_name_to_param_name = grad_name_to_param_name sparse_pair_map = collections.OrderedDict() for pair in self.origin_sparse_pairs + self.origin_dense_pairs: param, grad = pair sparse_pair_map[param.name] = str(param) sparse_pair_map[grad.name] = str(grad) self._var_slice_and_distribute() self._dispatcher() def get_param_grads(self): origin_program = self.origin_main_program def _get_params_grads(sparse_varnames): block = origin_program.global_block() dense_param_grads = [] sparse_param_grads = [] optimize_params = set() origin_var_dict = origin_program.global_block().vars role_id = int(core.op_proto_and_checker_maker.OpRole.Backward) for op in block.ops: if _is_opt_role_op(op): # delete clip op from opt_ops when run in Parameter Server mode if ( OP_NAME_SCOPE in op.all_attrs() and CLIP_OP_NAME_SCOPE in op.attr(OP_NAME_SCOPE) ): op._set_attr("op_role", role_id) continue if op.attr(OP_ROLE_VAR_ATTR_NAME): param_name = op.attr(OP_ROLE_VAR_ATTR_NAME)[0] grad_name = op.attr(OP_ROLE_VAR_ATTR_NAME)[1] if param_name not in optimize_params: optimize_params.add(param_name) param_grad = ( origin_var_dict[param_name], origin_var_dict[grad_name], ) if param_name in sparse_varnames: sparse_param_grads.append(param_grad) else: dense_param_grads.append(param_grad) return sparse_param_grads, dense_param_grads def _get_sparse_varnames(): varnames = [] for op in origin_program.global_block().ops: if ( op.type in SPARSE_OP_TYPE_DICT.keys() and op.attr('remote_prefetch') is True ): param_name = op.input(SPARSE_OP_TYPE_DICT[op.type])[0] varnames.append(param_name) return list(set(varnames)) sparse_varnames = _get_sparse_varnames() sparse_param_grads, dense_param_grads = _get_params_grads( sparse_varnames ) return sparse_param_grads, dense_param_grads def remove_var_pair_by_grad(self, var_name): for index, pair in enumerate(self.merged_variables_pairs): var = pair[0] var_grad = pair[1] if var_grad.merged_var.name == var_name: del self.merged_variables_pairs[index] for index, pair in enumerate(self.merged_dense_pairs): var = pair[0] var_grad = pair[1] if var_grad.merged_var.name == var_name: del self.merged_dense_pairs[index] return for index, pair in enumerate(self.merged_sparse_pairs): var = pair[0] var_grad = pair[1] if var_grad.merged_var.name == var_name: del self.merged_sparse_pairs[index] return print("Not find {} in self.merge_pairs".format(var_name)) def _is_opt_role_op(op): # NOTE : depend on oprole to find out whether this op is for # optimize op_maker = core.op_proto_and_checker_maker optimize_role = core.op_proto_and_checker_maker.OpRole.Optimize if op_maker.kOpRoleAttrName() in op.attr_names and int( op.all_attrs()[op_maker.kOpRoleAttrName()] ) == int(optimize_role): return True return False def _get_optimize_ops(_program): block = _program.global_block() opt_ops = [] for op in block.ops: if _is_opt_role_op(op): # delete clip op from opt_ops when run in Parameter Server mode if ( OP_NAME_SCOPE in op.all_attrs() and CLIP_OP_NAME_SCOPE in op.attr(OP_NAME_SCOPE) ): op._set_attr( "op_role", int(core.op_proto_and_checker_maker.OpRole.Backward), ) continue opt_ops.append(op) return opt_ops def _add_lr_decay_table_pass(main_program, compiled_config, lr_decay_steps): if hasattr(compiled_config.origin_main_program, 'lr_sheduler'): from paddle.optimizer.lr import LRScheduler assert isinstance( compiled_config.origin_main_program.lr_sheduler, LRScheduler ), "must be LRScheduler" ops = _get_optimize_ops(compiled_config.origin_main_program) lr_param_dict = _get_lr_param_dict(ops) ( lr_decay_main_program, lr_decay_startup_program, lr_name, ) = _get_lr_sheduler_program( compiled_config.origin_main_program.lr_sheduler, lr_param_dict, lr_decay_steps, ) compiled_config.add_tensor_table( "@LR_DECAY_COUNTER@", lr_name, lr_decay_startup_program, lr_decay_main_program, "GlobalStepTable", ) def _get_lr_param_dict(opt_ops): lr_param_dict = {} for op in opt_ops: lr_name = op.input("LearningRate")[0] param_name = op.input("Param")[0] if lr_name not in lr_param_dict: lr_param_dict[lr_name] = [] lr_param_dict[lr_name].append(param_name) return lr_param_dict def _get_lr_sheduler_program(lr_sheduler, lr_param_dict, lr_decay_steps): schedler_decay = [ 'NoamDecay', 'NaturalExpDecay', 'InverseTimeDecay', 'ExponentialDecay', ] from paddle.optimizer.lr import ( ExponentialDecay, InverseTimeDecay, NaturalExpDecay, NoamDecay, ) from paddle.static.learning_rate_scheduler import ( exponential_decay, inverse_time_decay, natural_exp_decay, noam_decay, ) decay_main_program = paddle.static.Program() decay_startup_program = paddle.static.Program() lr_name = "" if isinstance(lr_sheduler, ExponentialDecay): with paddle.static.program_guard( decay_main_program, decay_startup_program ): lr = exponential_decay(1.0, lr_decay_steps, lr_sheduler.gamma, True) lr_name = lr.name logging.warn( "ExponentialDecay is set, staircase = True, global learning rate decay step is [ %d ], Change decay steps as follow: \n" "\t strategy = paddle.distributed.fleet.DistributedStrategy() \n " "\t strategy.a_sync = True \n" "\t strategy.a_sync_configs= { 'lr_decay_steps' : YOUR_DECAY_STEP } \n" % lr_decay_steps ) elif isinstance(lr_sheduler, NoamDecay): with paddle.static.program_guard( decay_main_program, decay_startup_program ): lr = noam_decay(lr_sheduler.d_model, lr_sheduler.warmup_steps, 1.0) lr_name = lr.name logging.warn( "NoamDecay is set, warmup steps is [ %d ]" % lr_sheduler.warmup_steps ) elif isinstance(lr_sheduler, NaturalExpDecay): with paddle.static.program_guard( decay_main_program, decay_startup_program ): lr = natural_exp_decay(1.0, lr_decay_steps, lr_sheduler.gamma, True) lr_name = lr.name logging.warn( "NaturalExpDecay is set, staircase = True, global learning rate decay step is [ %d ], Change decay steps as follow: \n" "\t strategy = paddle.distributed.fleet.DistributedStrategy() \n " "\t strategy.a_sync = True \n" "\t strategy.a_sync_configs= { 'lr_decay_steps' : YOUR_DECAY_STEP } \n" % lr_decay_steps ) elif isinstance(lr_sheduler, InverseTimeDecay): with paddle.static.program_guard( decay_main_program, decay_startup_program ): lr = inverse_time_decay( 1.0, lr_decay_steps, lr_sheduler.gamma, True ) lr_name = lr.name logging.warn( "InverseTimeDecay is set, staircase = True, global learning rate decay step is [ %d ], Change decay steps as follow: \n" "\t strategy = paddle.distributed.fleet.DistributedStrategy() \n " "\t strategy.a_sync = True \n" "\t strategy.a_sync_configs= { 'lr_decay_steps' : YOUR_DECAY_STEP } \n" % lr_decay_steps ) else: raise ValueError( "Not supported current LearningRate strategy, please use follow decay strategy: {}".format( schedler_decay ) ) return decay_main_program, decay_startup_program, lr_name def _get_varname_parts(varname): # returns origin, blockid, trainerid orig_var_name = "" trainer_part = "" block_part = "" trainer_idx = varname.find(".trainer_") if trainer_idx >= 0: trainer_part = varname[trainer_idx + 1 :] else: trainer_idx = len(varname) block_index = varname.find(".block") if block_index >= 0: block_part = varname[block_index + 1 : trainer_idx] else: block_index = len(varname) orig_var_name = varname[0 : min(block_index, trainer_idx)] return orig_var_name, block_part, trainer_part def _orig_varname(varname): orig, _, _ = _get_varname_parts(varname) return orig