def upload_samples(self, module_id, samples_with_categories, sleep_if_throttled=True,
features_schema=None):
url = self.endpoint + module_id + '/samples/'
samples = []
for i, s in enumerate(samples_with_categories):
# if is multi-feature
if isinstance(s[0], dict):
sample_dict = {"features": s[0]}
elif isinstance(s[0], six.string_types):
sample_dict = {"text": s[0]}
else:
raise MonkeyLearnException('The sample must be a text in sample ' + str(i))
if (isinstance(s[1], int) or
(isinstance(s[1], list) and all(isinstance(c, int) for c in s[1]))):
sample_dict["category_id"] = s[1]
elif (isinstance(s[1], six.string_types) or
(isinstance(s[1], list) and all(isinstance(c, six.string_types) for c in s[1]))):
sample_dict["category_path"] = s[1]
elif s[1] is not None:
raise MonkeyLearnException('Invalid category value in sample ' + str(i))
if (len(s) > 2 and s[2] and (isinstance(s[2], six.string_types) or
(isinstance(s[2], list) and all(isinstance(c, six.string_types) for c in s[2])))):
sample_dict['tag'] = s[2]
samples.append(sample_dict)
data = {
'samples': samples
}
if features_schema:
data['features_schema'] = features_schema
response = self.make_request(url, 'POST', data, sleep_if_throttled)
self.handle_errors(response)
return MonkeyLearnResponse(response.json()['result'], [response])
def check_batch_limits(self, text_list, batch_size):
if batch_size > MAX_BATCH_SIZE or batch_size < MIN_BATCH_SIZE:
raise MonkeyLearnException('batch_size has to be between {0} and {1}'.format(
MIN_BATCH_SIZE, MAX_BATCH_SIZE))
if not text_list:
raise MonkeyLearnException('The text_list can\'t be empty.')
if '' in text_list:
raise MonkeyLearnException('You have an empty text in position {0} in text_list'.format(
text_list.index('')))
for i, text in enumerate(text_list):
if not isinstance(text, six.string_types):
raise MonkeyLearnException(
'Element in position {0} in text_list must be a string.'.format(i)
)
def upload_samples(self,
module_id,
samples_with_categories,
sleep_if_throttled=True):
url = self.endpoint + module_id + '/samples/'
samples = []
for i, s in enumerate(samples_with_categories):
if (isinstance(s[1], int)
or (isinstance(s[1], list)
and all(isinstance(c, int) for c in s[1]))):
sample_dict = {"text": s[0], "category_id": s[1]}
elif (isinstance(s[1], basestring)
or (isinstance(s[1], list)
and all(isinstance(c, basestring) for c in s[1]))):
sample_dict = {"text": s[0], "category_path": s[1]}
elif s[1] is None:
sample_dict = {"text": s[0]}
else:
raise MonkeyLearnException(
'Invalid category value in sample ' + str(i))
if (len(s) > 2 and s[2]
and (isinstance(s[2], basestring) or
(isinstance(s[2], list)
and all(isinstance(c, basestring) for c in s[2])))):
sample_dict['tag'] = s[2]
samples.append(sample_dict)
data = {'samples': samples}
response = self.make_request(url, 'POST', data, sleep_if_throttled)
self.handle_errors(response)
return MonkeyLearnResponse(response.json()['result'], [response])
def predict(self, module_id, sample_list=None, sandbox=False,
batch_size=DEFAULT_BATCH_SIZE, sleep_if_throttled=True):
try:
sample_list = list(sample_list)
except TypeError:
raise MonkeyLearnException('The sample_list can\'t be None.')
self.check_batch_limits(sample_list, batch_size)
url = self.endpoint + module_id + '/predict/'
url_params = {}
if sandbox:
url_params['sandbox'] = 1
if url_params:
url += '?{}'.format(urlencode(url_params))
res = []
responses = []
for i in range(0, len(sample_list), batch_size):
data = {
'text_list': sample_list[i:i + batch_size]
}
response = self.make_request(url, 'POST', data, sleep_if_throttled)
self.handle_errors(response)
responses.append(response)
res.extend(response.json()['result'])
return MonkeyLearnResponse(res, responses)
def handle_errors(self, response):
if not response.ok:
try:
res = response.json()
except ValueError:
response.raise_for_status()
raise MonkeyLearnException(json.dumps(res['detail']))
def run(self, module_id, data, sandbox=False, sleep_if_throttled=True):
if not isinstance(data, dict):
raise MonkeyLearnException('The data parameter must be a dictionary')
url = self.endpoint + module_id + '/run/'
if sandbox:
url += '?sandbox=1'
response = self.make_request(url, 'POST', data, sleep_if_throttled)
self.handle_errors(response)
return MonkeyLearnResponse(response.json()['result'], [response])
def classify(self,
module_id,
sample_list=None,
sandbox=False,
batch_size=DEFAULT_BATCH_SIZE,
sleep_if_throttled=True,
debug=False,
text_list=None):
if text_list:
warnings.warn(
"The text_list parameter will be deprecated in future versions. Please use sample_list."
)
sample_list = text_list
try:
sample_list = list(sample_list)
except TypeError:
raise MonkeyLearnException('The sample_list can\'t be None.')
self.check_batch_limits(sample_list, batch_size)
url = self.endpoint + module_id + '/classify/'
url_params = {}
if sandbox:
url_params['sandbox'] = 1
if debug:
url_params['debug'] = 1
if url_params:
url += '?{}'.format(urlencode(url_params))
res = []
responses = []
for i in range(0, len(sample_list), batch_size):
# if is multi feature
if isinstance(sample_list[0], dict):
data = {'sample_list': sample_list[i:i + batch_size]}
else:
data = {'text_list': sample_list[i:i + batch_size]}
response = self.make_request(url, 'POST', data, sleep_if_throttled)
self.handle_errors(response)
responses.append(response)
res.extend(response.json()['result'])
return MonkeyLearnResponse(res, responses)
def upload_samples(self, module_id, samples_to_upload, sleep_if_throttled=True):
url = self.endpoint + module_id + '/samples/'
samples = []
for i, s in enumerate(samples_to_upload):
if isinstance(s[0], six.string_types):
sample_dict = {"text": s[0]}
else:
raise MonkeyLearnException('The sample must be a text in sample ' + str(i))
if (len(s) > 1 and s[1] and (isinstance(s[1], six.string_types) or
(isinstance(s[1], list) and all(isinstance(c, six.string_types) for c in s[1])))):
sample_dict['tag'] = s[1]
samples.append(sample_dict)
data = {
'samples': samples
}
response = self.make_request(url, 'POST', data, sleep_if_throttled)
self.handle_errors(response)
return MonkeyLearnResponse(response.json()['result'], [response])