def make_serving_input_fn(self, tf_transform_output):
""" Estimator input function generator for model serving.
:param tf_transform_output: tf.Transform graph output wrapper.
:return: Estimator input function for serving (prediction).
"""
data_formatter = DataFormatter()
def serving_input_fn():
"""
inputs : supported features
inputs_ext: all features
"""
inputs, inputs_ext = {}, {}
# Used input features
for key in data_formatter.FEATURES:
placeholder = tf.placeholder(
shape=[None], dtype=data_formatter.get_tf_dtype(key))
inputs[key] = placeholder
inputs_ext[key] = placeholder
transformed_features = tf_transform_output.transform_raw_features(
inputs)
return tf.estimator.export.ServingInputReceiver(
transformed_features, inputs_ext)
return serving_input_fn
def get_dict_tweets_list(self, limit, query, url):
data_formatter = DataFormatter()
dict_tweets_list = []
i = 0
while i < limit:
if (i == 0):
base_url = url.format(q=query, pos="")
else:
base_url = url.format(q=query, pos=next_position)
url_data = requests.get(base_url, headers=self.HEADER)
if url_data.status_code != 200:
break
content = json.loads(url_data.text)
next_position = content['min_position']
html_data = content['items_html']
soup = BeautifulSoup(html_data)
tweet_blocks = soup.select('.tweet')
for tweet_block in tweet_blocks:
if (i < limit):
dict_tweets_list.append(
data_formatter.get_dict_one_tweet_data(tweet_block))
i += 1
else:
break
if content['has_more_items'] == False:
break
return dict_tweets_list
def make_serving_input_fn(self, tf_transform_output):
""" Estimator input function generator for model serving.
:param tf_transform_output: tf.Transform graph output wrapper.
:return: Estimator input function for serving (prediction).
"""
data_formatter = DataFormatter()
def serving_input_fn():
"""
inputs : supported features
inputs_ext: all features
"""
inputs, inputs_ext = {}, {}
# Used input features
for key in data_formatter.FEATURES:
placeholder = tf.placeholder(
shape=[None], dtype=data_formatter.get_tf_dtype(key))
inputs[key] = placeholder
inputs_ext[key] = placeholder
transformed_features = tf_transform_output.transform_raw_features(inputs)
# todo: try RNN List features
tensors = []
tensors.append(transformed_features["close_b20"])
tensors.append(transformed_features["close_b19"])
tensors.append(transformed_features["close_b18"])
tensors.append(transformed_features["close_b17"])
tensors.append(transformed_features["close_b16"])
tensors.append(transformed_features["close_b15"])
tensors.append(transformed_features["close_b14"])
tensors.append(transformed_features["close_b13"])
tensors.append(transformed_features["close_b12"])
tensors.append(transformed_features["close_b11"])
tensors.append(transformed_features["close_b10"])
tensors.append(transformed_features["close_b9"])
tensors.append(transformed_features["close_b8"])
tensors.append(transformed_features["close_b7"])
tensors.append(transformed_features["close_b6"])
tensors.append(transformed_features["close_b5"])
tensors.append(transformed_features["close_b4"])
tensors.append(transformed_features["close_b3"])
tensors.append(transformed_features["close_b2"])
tensors.append(transformed_features["close_b1"])
tensors.append(transformed_features["close_b0"])
tensors_concat = tf.stack(tensors, axis=1)
return tf.estimator.export.ServingInputReceiver({"closes": tensors_concat}, inputs_ext)
return serving_input_fn
def __init__(self):
self.data_formatter = DataFormatter()
# Classification and regresion target definition
self.CLASSIF_TARGETS = self.data_formatter.TARGETS
def __init__(self):
self.data_formatter = DataFormatter()
self.CLASSIF_TARGETS = self.data_formatter.TARGETS
def __init__(self):
self.data_formatter = DataFormatter()
# from trackml.score import score_event
# import pdb
# import pandas as pd
# import csv
from data_formatter import DataFormatter
## Load Data ##
path_to_dataset = "../../Data/train_100_events/"
event_path = "event000001052"
model_name = "identity.keras"
hits, cells, particles, truth = load_event(path_to_dataset + event_path)
# Get the sorted tracks
formatter = DataFormatter()
true_tracks, hit_tracks = formatter.getSortedTracks(particles, truth, hits)
## Load Predicted Seeds ##
seed_file = open("SeedCandidates.txt", "r")
our_tracks = []
seed_hits = []
np_hits = np.asarray(hits)
for seed_id in seed_file:
seed_id = int(float(seed_id.strip()))
seed_hit = np_hits[np_hits[:, 0] == seed_id][0]
our_tracks.append([int(seed_hit[0])])
seed_hits.append(seed_hit)
print("\nStarting with " + str(len(seed_hits)) + " seed hits")
def __init__(self):
self.schema = DataFormatter()
