AI-MXNet
view release on metacpan - search on metacpan
view release on metacpan or search on metacpan
lib/AI/MXNet/Executor/Group.pm view on Meta::CPAN
has [qw/output_layouts label_layouts arg_names aux_names
batch_size slices execs data_arrays
label_arrays param_arrays grad_arrays aux_arrays
data_layouts shared_data_arrays input_grad_arrays
_default_execs state_arrays/
] => (is => 'rw', init_arg => undef);
package AI::MXNet::DataParallelExecutorGroup;
use Mouse;
use AI::MXNet::Base;
use List::Util qw(sum);
=head1 DESCRIPTION
DataParallelExecutorGroup is a group of executors that lives on a group of devices.
This is a helper class used to implement data parallelization. Each mini-batch will
be split and run on the devices.
Parameters for constructor
----------
symbol : AI::MXNet::Symbol
The common symbolic computation graph for all executors.
contexts : ArrayRef[AI::MXNet::Context]
A array ref of contexts.
workload : ArrayRef[Num]
If not undef, could be an array ref of numbers that specify the workload to be assigned
to different context. Larger number indicate heavier workload.
data_shapes : ArrayRef[NameShape|AI::MXNet::DataDesc]
Should be a array ref of [name, shape] array refs, for the shapes of data. Note the order is
important and should be the same as the order that the `DataIter` provide the data.
label_shapes : Maybe[ArrayRef[NameShape|AI::MXNet::DataDesc]]
Should be a array ref of [$name, $shape] array refs, for the shapes of label. Note the order is
important and should be the same as the order that the `DataIter` provide the label.
param_names : ArrayRef[Str]
A array ref of strings, indicating the names of parameters (e.g. weights, filters, etc.)
in the computation graph.
for_training : Bool
Indicate whether the executors should be bind for training. When not doing training,
the memory for gradients will not be allocated.
inputs_need_grad : Bool
Indicate whether the gradients for the input data should be computed. This is currently
not used. It will be useful for implementing composition of modules.
shared_group : AI::MXNet::DataParallelExecutorGroup
Default is undef. This is used in bucketing. When not undef, it should be a executor
group corresponding to a different bucket. In other words, it will correspond to a different
symbol with the same set of parameters (e.g. unrolled RNNs with different lengths).
In this case the memory regions of the parameters will be shared.
logger : Logger
Default is AI::MXNet::Logging->get_logger.
fixed_param_names: Maybe[ArrayRef[Str]]
Indicate parameters to be fixed during training. Parameters in this array ref will not allocate
space for gradient, nor do gradient calculation.
grad_req : ArrayRef[GradReq]|HashRef[GradReq]|GradReq
Requirement for gradient accumulation. Can be 'write', 'add', or 'null'
(default to 'write').
Can be specified globally (str) or for each argument (array ref, hash ref).
state_names: Maybe[ArrayRef[Str]]
=cut
has 'symbol' => (is => 'ro', isa => 'AI::MXNet::Symbol', required => 1);
has 'contexts' => (is => 'ro', isa => 'ArrayRef[AI::MXNet::Context]', required => 1);
has 'workload' => (is => 'ro', isa => 'ArrayRef[Num]', default => sub { [] });
has 'data_shapes' => (is => 'rw', isa => 'ArrayRef[NameShape|AI::MXNet::DataDesc]', required => 1);
has 'label_shapes' => (is => 'rw', isa => 'Maybe[ArrayRef[NameShape|AI::MXNet::DataDesc]]');
has 'param_names' => (is => 'ro', isa => 'ArrayRef[Str]', required => 1);
has 'for_training' => (is => 'ro', isa => 'Bool', required => 1);
has 'inputs_need_grad' => (is => 'ro', isa => 'Bool', default => 0);
has 'shared_group' => (is => 'ro', isa => 'Maybe[AI::MXNet::DataParallelExecutorGroup]');
has 'logger' => (is => 'ro', default => sub { AI::MXNet::Logging->get_logger });
has 'fixed_param_names' => (is => 'rw', isa => 'Maybe[ArrayRef[Str]]');
has 'state_names' => (is => 'rw', isa => 'Maybe[ArrayRef[Str]]');
has 'grad_req' => (is => 'rw', isa => 'ArrayRef[GradReq]|HashRef[GradReq]|GradReq', default=>'write');
has '_p' => (is => 'rw', init_arg => undef);
sub BUILD
{
my $self = shift;
my $p = AI::MXNet::DataParallelExecutorGroup::_private->new;
$p->arg_names($self->symbol->list_arguments);
$p->aux_names($self->symbol->list_auxiliary_states);
$p->execs([]);
$self->_p($p);
$self->grad_req('null') if not $self->for_training;
$self->fixed_param_names([]) unless defined $self->fixed_param_names;
$self->state_names([]) unless defined $self->state_names;
my $data_shapes = [];
for my $d (@{ $self->data_shapes })
{
$d = AI::MXNet::DataDesc->new(name => $d->[0], shape => $d->[1])
unless blessed $d;
push @{ $data_shapes }, $d;
}
$self->data_shapes($data_shapes);
if(defined $self->label_shapes)
{
my $label_shapes = [];
for my $l (@{ $self->label_shapes })
{
$l = AI::MXNet::DataDesc->new(name => $l->[0], shape => $l->[1])
unless blessed $l;
push @{ $label_shapes }, $l;
}
$self->label_shapes($label_shapes);
}
my %data_names = map { $_->name => 1 } @{ $self->data_shapes };
my %param_names = map { $_ => 1 } @{ $self->param_names };
my %fixed_param_names = map { $_ => 1 } @{ $self->fixed_param_names };
my %grad_req;
if(not ref $self->grad_req)
{
for my $k (@{ $self->_p->arg_names })
{
if(exists $param_names{ $k })
{
$grad_req{$k} = exists $fixed_param_names{ $k } ? 'null' : $self->grad_req;
}
elsif(exists $data_names{ $k })
{
$grad_req{$k} = $self->inputs_need_grad ? $self->grad_req : 'null';
}
else
{
$grad_req{$k} = 'null';
}
}
}
view all matches for this distributionview release on metacpan - search on metacpan
( run in 2.621 seconds using v1.00-cache-2.02-grep-82fe00e-cpan-72ae3ad1e6da )