def eval_classification(model, sessinit, dataflow):
pred_config = PredictConfig(
model=model,
session_init=sessinit,
input_names=['input', 'label'],
output_names=['wrong-top1','bn5/output:0','conv5/output:0']
)
acc1 = RatioCounter()
pred = FeedfreePredictor(pred_config, StagingInput(QueueInput(dataflow), device='/gpu:0'))
for _ in tqdm.trange(dataflow.size()):
top1,afbn5,beforbn5= pred()
dic ={}
dic['bn5/output:0']=afbn5
dic['conv5/output:0']=beforbn5
batch_size = top1.shape[0]
acc1.feed(top1.sum(), batch_size)
dir = logger.get_logger_dir()
fname = os.path.join(
dir, 'afbn5-{}.npz'.format(int(time.time())))
np.savez(fname, **dic)
print("Top1 Error: {}".format(acc1.ratio))
def pred_config(self, args, df, callbacks) -> TrainConfig:
return TrainConfig(
model=self.train_model(args),
data=StagingInput(QueueInput(df)),
callbacks=callbacks,
max_epoch=args.epochs,
steps_per_epoch=args.steps,
session_init=SaverRestore(args.load) if args.load else None,
)
def test(net,
session_init,
val_dataflow,
do_calc_flops=False,
extended_log=False):
"""
Main test routine.
Parameters:
----------
net : obj
Model.
session_init : SessionInit
Session initializer.
do_calc_flops : bool, default False
Whether to calculate count of weights.
extended_log : bool, default False
Whether to log more precise accuracy values.
"""
pred_config = PredictConfig(
model=net,
session_init=session_init,
input_names=["input", "label"],
output_names=["wrong-top1", "wrong-top5"]
)
err_top1 = RatioCounter()
err_top5 = RatioCounter()
tic = time.time()
pred = FeedfreePredictor(pred_config, StagingInput(QueueInput(val_dataflow), device="/gpu:0"))
for _ in tqdm.trange(val_dataflow.size()):
err_top1_val, err_top5_val = pred()
batch_size = err_top1_val.shape[0]
err_top1.feed(err_top1_val.sum(), batch_size)
err_top5.feed(err_top5_val.sum(), batch_size)
err_top1_val = err_top1.ratio
err_top5_val = err_top5.ratio
if extended_log:
logging.info("Test: err-top1={top1:.4f} ({top1})\terr-top5={top5:.4f} ({top5})".format(
top1=err_top1_val, top5=err_top5_val))
else:
logging.info("Test: err-top1={top1:.4f}\terr-top5={top5:.4f}".format(
top1=err_top1_val, top5=err_top5_val))
logging.info("Time cost: {:.4f} sec".format(
time.time() - tic))
if do_calc_flops:
calc_flops(model=net)
def eval_on_ILSVRC12(model, sessinit, dataflow):
pred_config = PredictConfig(model=model,
session_init=sessinit,
input_names=['input', 'label'],
output_names=['wrong-top1', 'wrong-top5'])
acc1, acc5 = RatioCounter(), RatioCounter()
# This does not have a visible improvement over naive predictor,
# but will have an improvement if image_dtype is set to float32.
pred = FeedfreePredictor(
pred_config, StagingInput(QueueInput(dataflow), device='/gpu:0'))
for _ in tqdm.trange(dataflow.size()):
top1, top5 = pred()
batch_size = top1.shape[0]
acc1.feed(top1.sum(), batch_size)
acc5.feed(top5.sum(), batch_size)
print("Top1 Error: {}".format(acc1.ratio))
print("Top5 Error: {}".format(acc5.ratio))
def test(net,
session_init,
val_dataflow,
do_calc_flops=False,
extended_log=False):
pred_config = PredictConfig(
model=net,
session_init=session_init,
input_names=['input', 'label'],
output_names=['wrong-top1', 'wrong-top5']
)
err_top1 = RatioCounter()
err_top5 = RatioCounter()
tic = time.time()
pred = FeedfreePredictor(pred_config, StagingInput(QueueInput(val_dataflow), device='/gpu:0'))
# import tensorflow as tf
# summ_writer = tf.summary.FileWriter("/home/semery/projects/imgclsmob_data/gl-squeezenet_v1_1/", pred._sess.graph)
for _ in tqdm.trange(val_dataflow.size()):
err_top1_val, err_top5_val = pred()
batch_size = err_top1_val.shape[0]
err_top1.feed(err_top1_val.sum(), batch_size)
err_top5.feed(err_top5_val.sum(), batch_size)
# print("err_top1_val={}".format(err_top1_val.sum() / batch_size))
# print("err_top5_val={}".format(err_top5_val.sum() / batch_size))
err_top1_val = err_top1.ratio
err_top5_val = err_top5.ratio
if extended_log:
logging.info('Test: err-top1={top1:.4f} ({top1})\terr-top5={top5:.4f} ({top5})'.format(
top1=err_top1_val, top5=err_top5_val))
else:
logging.info('Test: err-top1={top1:.4f}\terr-top5={top5:.4f}'.format(
top1=err_top1_val, top5=err_top5_val))
logging.info('Time cost: {:.4f} sec'.format(
time.time() - tic))
if do_calc_flops:
calc_flops(model=net)
def test(net,
session_init,
val_dataflow,
do_calc_flops=False,
extended_log=False):
pred_config = PredictConfig(
model=net,
session_init=session_init,
input_names=["input", "label"],
output_names=["wrong-top1", "wrong-top5"]
)
err_top1 = RatioCounter()
err_top5 = RatioCounter()
tic = time.time()
pred = FeedfreePredictor(pred_config, StagingInput(QueueInput(val_dataflow), device="/gpu:0"))
for _ in tqdm.trange(val_dataflow.size()):
err_top1_val, err_top5_val = pred()
batch_size = err_top1_val.shape[0]
err_top1.feed(err_top1_val.sum(), batch_size)
err_top5.feed(err_top5_val.sum(), batch_size)
err_top1_val = err_top1.ratio
err_top5_val = err_top5.ratio
if extended_log:
logging.info("Test: err-top1={top1:.4f} ({top1})\terr-top5={top5:.4f} ({top5})".format(
top1=err_top1_val, top5=err_top5_val))
else:
logging.info("Test: err-top1={top1:.4f}\terr-top5={top5:.4f}".format(
top1=err_top1_val, top5=err_top5_val))
logging.info("Time cost: {:.4f} sec".format(
time.time() - tic))
if do_calc_flops:
calc_flops(model=net)
def eval_classification(model, sessinit, dataflow):
"""
Eval a classification model on the dataset. It assumes the model inputs are
named "input" and "label", and contains "wrong-top1" and "wrong-top5" in the graph.
"""
pred_config = PredictConfig(model=model,
session_init=sessinit,
input_names=['input', 'label'],
output_names=['wrong-top1', 'wrong-top5'])
acc1, acc5 = RatioCounter(), RatioCounter()
# This does not have a visible improvement over naive predictor,
# but will have an improvement if image_dtype is set to float32.
pred = FeedfreePredictor(
pred_config, StagingInput(QueueInput(dataflow), device='/gpu:0'))
for _ in tqdm.trange(dataflow.size()):
top1, top5 = pred()
batch_size = top1.shape[0]
acc1.feed(top1.sum(), batch_size)
acc5.feed(top5.sum(), batch_size)
print("Top1 Error: {}".format(acc1.ratio))
print("Top5 Error: {}".format(acc5.ratio))
