from __future__ import print_function import framework from framework import Program, default_main_program, Parameter, Variable import optimizer from layer_helper import LayerHelper from distributed_spliter import * import math class VarBlock: def __init__(self, varname, offset, size): self.varname = varname # NOTE: real offset is offset * size self.offset = offset self.size = size def __str__(self): return "%s:%d:%d" % (self.varname, self.offset, self.size) def split_dense_variable(var_list, pserver_count, min_block_size=1024, max_block_size=1048576): """ We may need to split dense tensor to one or several 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 mininum block size is 1024. The max block size is used to prevent too large block that may causing send error. """ blocks = [] for var in var_list: split_count = pserver_count var_numel = reduce(lambda x, y: x * y, var.shape) 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 < pserver_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 align split_count = int(math.ceil(var_numel / float(block_size))) for block_id in xrange(split_count): curr_block_size = min(block_size, var_numel - ( (block_id) * block_size)) block = VarBlock(var.name, block_id, curr_block_size) blocks.append(str(block)) return blocks class DistributeTranspiler: def transpile(self, optimize_ops, params_grads, program=None, pservers="127.0.0.1:6174", trainers=1, split_method=round_robin): """ Transpile the program to a distributed data-parallelism programs. The main_program will be transform to use a remote parameter server to do parameter optimization. And the optimization graph will be put in to a parameter server program. Use different methods to split trainable varialbles to different parameter servers. :param optimize_ops: op list of optimization, should be the return value of Optimizer.minimize :type optimize_ops: list :param program: program to optimize, default default_main_program :param pservers: parameter server endpoints like "m1:6174,m2:6174" :type pservers: string :return: return a list of programs """ assert (callable(split_method)) if program is None: program = default_main_program() self.program = program self.trainers = trainers self.optimize_ops = optimize_ops # steps to transpile: # 1. split variable to multiple blocks, align by product(dim[1:]) (width). # 2. modify trainer program add split_op to each Grad. # 3. append send_op to trainer. # 4. append concat_op to trainer to update local weights. # 5. create new program as parameter server. # 5. create parameter server program by split_method generated endpoint->VarBlock # 6. run compile time infershape for parameter server program pserver_endpoints = pservers.split(",") # step1 param_list = [pg[0] for pg in params_grads] grad_list = [pg[1] for pg in params_grads] # TODO: add split selected rows support grad_blocks = split_dense_variable(grad_list, len(pserver_endpoints)) param_blocks = split_dense_variable(param_list, len(pserver_endpoints)) # step2 grad_var_mapping = self._append_split_op(program, grad_blocks) # step3 send_inputs = [] send_outputs = [] for _, splited in grad_var_mapping.iteritems(): send_inputs.extend(splited) param_var_mapping = self._create_vars_from_blocklist(program, param_blocks) for _, splited in param_var_mapping.iteritems(): send_outputs.extend(splited) # let send_op know which endpoint to send which var, eplist is of the same # order of send_inputs. eplist = split_method(send_inputs, pserver_endpoints) send_op = program.global_block().append_op( type="send", inputs={"X": send_inputs}, outputs={"Out": send_outputs}, attrs={"endpoints": pserver_endpoints, "epmap": eplist}) # step4 for varname, splited_var in param_var_mapping.iteritems(): orig_param = program.global_block().vars[varname] concat = program.global_block().append_op( type="concat", inputs={"X": send_outputs}, outputs={"Out": orig_param}, attrs={"axis": 0}) def _create_vars_from_blocklist(self, program, block_list): block_map = dict() var_mapping = dict() for block_str in block_list: varname, offset, size = block_str.split(":") if not block_map.has_key(varname): block_map[varname] = [] block_map[varname].append((long(offset), long(size))) for varname, splited in block_map.iteritems(): orig_var = program.global_block().vars[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:]) var_list = [] for i, block in enumerate(splited): size = block[1] rows = size / orig_dim1_flatten splited_shape = [rows] if len(orig_shape) >= 2: splited_shape.extend(orig_shape[1:]) print("block, splited shape:", block, splited_shape) var = program.global_block().create_var( name="%s.block%d" % (varname, i), psersistable=False, dtype=orig_var.dtype, shape=splited_shape) # flattend splited var var_list.append(var) var_mapping[varname] = var_list return var_mapping def _clone_param(self, block, v): assert isinstance(v, Parameter) new_p = Parameter( block=block, shape=v.shape, dtype=v.dtype, type=v.type, lod_level=v.lod_level, stop_gradient=v.stop_gradient, trainable=v.trainable, optimize_attr=v.optimize_attr, regularizer=v.regularizer, name=v.name) block.vars[new_p.name] = new_p def _clone_var(self, block, var): assert isinstance(var, Variable) return block.create_var( name=var.name, shape=var.shape, dtype=var.dtype, type=var.type, lod_level=var.lod_level, persistable=var.persistable) def _append_split_op(self, program, gradblocks): var_mapping = self._create_vars_from_blocklist(program, gradblocks) for varname, splited_vars in var_mapping.iteritems(): if len(splited_vars) == 1: continue orig_var = program.global_block().vars[varname] sections = [] for v in splited_vars: sections.append(v.shape[0]) program.global_block().append_op( type="split", inputs={"X": orig_var}, outputs={"Out": splited_vars}, attrs={"sections": sections} # assume split evenly ) return var_mapping def get_trainer_program(self): # remove optimize ops and add a send op to main_program self.program.global_block().delete_ops(self.optimize_ops) return self.program def _create_var_for_trainers(self, block, var, trainers): var_list = [] for i in xrange(trainers): var_each = block.create_var( name="%s.trainer_%d" % (var.name, i), psersistable=var.persistable, dtype=var.dtype, shape=var.shape) var_list.append(var_each) return var_list def get_pserver_program(self, endpoint, optimize_ops): pserver_program = Program() for v in self.param_grad_map[endpoint]["params"]: self._clone_param(pserver_program.global_block(), v) optimize_sub_program = Program() grad_var_names = [ var.name for var in self.param_grad_map[endpoint]["grads"] ] for opt_op in optimize_ops: for _, var in opt_op.inputs.iteritems(): # NOTE: append operators to merge gradients from multiple # trainers. If trainers == 1, this is not needed. if self.trainers > 1 and var.name in grad_var_names: vars2merge = self._create_var_for_trainers( optimize_sub_program.global_block(), var, self.trainers) merged_var = optimize_sub_program.global_block().create_var( name=var.name, persistable=var.persistable, dtype=var.dtype, shape=var.shape) optimize_sub_program.global_block().append_op( type="sum", inputs={"X": vars2merge}, outputs={"Out": merged_var}) optimize_sub_program.global_block().append_op( type="scale", inputs={"X": merged_var}, outputs={"Out": merged_var}, attrs={"scale": 1.0 / float(self.trainers)}) else: optimize_sub_program.global_block().create_var( name=var.name, persistable=var.persistable, dtype=var.dtype, shape=var.shape) if opt_op.inputs.has_key("Grad"): if opt_op.inputs["Grad"].name in grad_var_names: optimize_sub_program.global_block().append_op( type=opt_op.type, inputs=opt_op.inputs, outputs=opt_op.outputs, attrs=opt_op.attrs) else: optimize_sub_program.global_block().append_op( type=opt_op.type, inputs=opt_op.inputs, outputs=opt_op.outputs, attrs=opt_op.attrs) pserver_program.global_block().append_op( type="recv", inputs={"RX": self.param_grad_map[endpoint]["grads"]}, # grads to recv outputs={}, attrs={ "OptimizeProgram": optimize_sub_program.desc, "endpoint": endpoint, "ParamList": [p.name for p in self.param_grad_map[endpoint]["params"]], "GradList": [p.name for p in self.param_grad_map[endpoint]["grads"]], "Trainers": self.trainers }) pserver_program.sync_with_cpp() return pserver_program