def transformer_factory(request):
def set_and_get_factory(transformer_name):
factory = ngt.make_transformer_factory(transformer_name)
ngt.set_transformer_factory(factory)
return factory
name = request.config.getoption("--transformer")
yield set_and_get_factory(name)
# Reset transformer factory to default
ngt.set_transformer_factory(ngt.make_transformer_factory("ngcpu"))
self.ngraph_backend = "INTERPRETER"
super(PybindINTERPRETERTransformer, self).__init__(**kwargs)
class PybindGPUTransformer(PybindTransformer):
"""
Transformer for ngraph c++ with gpu backend.
"""
transformer_name = "nggpu"
def __init__(self, **kwargs):
self.ngraph_backend = "GPU"
super(PybindGPUTransformer, self).__init__(**kwargs)
class PybindARGONTransformer(PybindTransformer):
"""
Transformer for ngraph c++ with argon backend.
"""
transformer_name = "ngargon"
def __init__(self, **kwargs):
self.ngraph_backend = "ARGON"
super(PybindARGONTransformer, self).__init__(**kwargs)
set_transformer_factory(
make_transformer_factory(PybindCPUTransformer.transformer_name))
restore_eval_function = transformer.add_computation(eval_computation)
weight_saver.setup_restore(transformer=transformer,
computation=eval_computation,
filename=args.inference)
# Restore weight
weight_saver.restore()
# Calculate losses
eval_losses = loop_eval(valid_set, input_ph, metric_names,
restore_eval_function, en_top5)
# Print statistics
print("From restored weights: Test Avg loss:{tcost}".format(
tcost=eval_losses))
exit()
# Training the network by calling transformer
with closing(ngt.make_transformer_factory(args.backend)()) as transformer:
# Trainer
train_function = transformer.add_computation(train_computation)
# Inference
eval_function = transformer.add_computation(eval_computation)
# Set Saver for saving weights
weight_saver.setup_save(transformer=transformer,
computation=train_computation)
# Resume weights for training from a checkpoint
if args.resume is not None:
weight_saver.setup_restore(transformer=transformer,
computation=train_computation,
filename=args.resume)
weight_saver.restore()
# Progress bar
tpbar = tqdm(unit="batches", ncols=100, total=args.num_iterations)
def make_and_set_transformer_factory(self, args):
factory = ngt.make_transformer_factory(args.backend)
ngt.set_transformer_factory(factory)
def set_and_get_factory(transformer_name):
factory = ngt.make_transformer_factory(transformer_name)
ngt.set_transformer_factory(factory)
return factory
