#
# Copyright 2018 Analytics Zoo Authors.
#
# 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.
#
from zoo.tfpark.tfnet import TFNet
from zoo.tfpark.tf_optimizer import BigDLMetric, TFModel
from zoo.pipeline.api.keras import metrics as zmetrics
[docs]def to_bigdl_metric(metric):
metric = metric.lower()
if metric == "accuracy" or metric == "acc":
return zmetrics.Accuracy()
elif metric == "top5accuracy" or metric == "top5acc":
return zmetrics.Top5Accuracy()
elif metric == "mae":
from bigdl.optim.optimizer import MAE
return MAE()
elif metric == "auc":
return zmetrics.AUC()
elif metric == "treennaccuracy":
from bigdl.optim.optimizer import TreeNNAccuracy
return TreeNNAccuracy()
else:
raise TypeError("Unsupported metric: %s" % metric)
[docs]def evaluate_string_metrics(*,
sess,
string_metrics,
dataset,
inputs,
targets=None,
outputs=None,
loss=None,
):
metrics = {}
for i, metric in enumerate(string_metrics):
if metric == "loss":
assert loss is not None, "loss tensor should not be None if one of the metrics is loss"
metrics["loss"] = loss
else:
assert outputs is not None, "outputs should not be None if non loss metrics exists"
assert targets is not None, "targets should not be None if non loss metrics exists"
method = to_bigdl_metric(metric)
metrics[metric] = BigDLMetric(method,
outputs,
targets)
result = evaluate_metrics(inputs, sess, dataset, metrics)
return result
[docs]def evaluate_metrics(inputs, sess, dataset, metrics):
import tensorflow as tf
if dataset.batch_per_thread > 0:
batch_size = dataset.batch_per_thread * dataset.get_num_partitions()
else:
batch_size = dataset.batch_size
real_batch_size = tf.shape(inputs[0])[0]
outputs, eval_methods = TFModel._process_metrics(inputs[0].graph,
metrics=metrics,
real_batch_size=real_batch_size)
tfnet = TFNet.from_session(sess, inputs=inputs, outputs=outputs)
results = tfnet.evaluate(dataset, batch_size, eval_methods)
final_result = dict([(r.method, r.result) for r in results])
return final_result