def call_model_cdsw(self, record):
"""
Not Implemented - currently performs 42% slower than call_model.
"""
response = cdsw.call_model(
model_access_key=self.deployment_details["model_access_key"],
ipt={"record": record},
)
return record["id"], response["response"]["uuid"]
# Create an array of model responses.
response_labels_sample = []
# Make 250 calls to the model with increasing error
percent_counter = 0
percent_max = len(df_sample_clean)
for record in json.loads(df_sample_clean.to_json(orient='records')):
print("Added {} records".format(percent_counter)) if (percent_counter %
50 == 0) else None
percent_counter += 1
no_churn_record = copy.deepcopy(record)
no_churn_record.pop('customerID')
no_churn_record.pop('Churn')
# **note** this is an easy way to interact with a model in a script
response = cdsw.call_model(latest_model["accessKey"], no_churn_record)
response_labels_sample.append({
"uuid":
response["response"]["uuid"],
"final_label":
churn_error(record["Churn"], percent_counter / percent_max),
"response_label":
response["response"]["prediction"]["probability"] >= 0.5,
"timestamp_ms":
int(round(time.time() * 1000))
})
# The "ground truth" loop adds the updated actual label value and an accuracy measure
# every 100 calls to the model.
for index, vals in enumerate(response_labels_sample):
print("Update {} records".format(index)) if (index % 50 == 0) else None
]].to_numpy()[0]
try:
input_data[1] = int(input_data[1])
input_data[4] = int(input_data[4])
input_data[5] = int(input_data[5])
input_data[6] = int(input_data[6])
input_data[9] = int(input_data[9])
input_data_string = ""
for record in input_data[:-1]:
input_data_string = input_data_string + str(record) + ","
input_data_string = input_data_string[:-1]
response = cdsw.call_model(latest_model["accessKey"],
{"feature": input_data_string})
predicted_result.append(
response["response"]["prediction"]["prediction"])
actual_result.append(input_data[-1:][0])
cdsw.track_delayed_metrics({"actual_result": input_data[-1:][0]},
response["response"]["uuid"])
print(str(i) + " adding " + input_data_string)
except:
print("invalid row")
time.sleep(0.2)
end_time_ms = int(math.floor(time.time() * 1000))
accuracy = classification_report(actual_result,
predicted_result,
output_dict=True)['accuracy']
cdsw.track_aggregate_metrics({"accuracy": accuracy},
# This is the input data for which we want to make predictions.
# Ground truth is generally not yet known at prediction time.
score_x = iris.data[:test_size, 2].reshape(-1, 1) # Petal length
# Record the current time so we can retrieve the metrics
# tracked for these calls.
start_timestamp_ms=int(round(time.time() * 1000))
print("output type")
type(output)
uuids = []
predictions = []
for i in range(len(score_x)):
output = cdsw.call_model(model_access_key, {"petal_length": score_x[i][0]}, api_key)
#Record the UUID of each prediction for correlation with ground truth.
uuids.append(output["response"]["uuid"])
predictions.append(output["response"]["prediction"])
print(output)
# Record the current time.
end_timestamp_ms=int(round(time.time() * 1000))
# We can now use the read_metrics function to read the metrics we just
# generated into the current session, by querying by time window.
data = cdsw.read_metrics(model_deployment_crn=model_deployment_crn,
start_timestamp_ms=start_timestamp_ms,
end_timestamp_ms=end_timestamp_ms)
# Run Similation to make 1000 calls to the model with increasing error
percent_counter = 0
percent_max = len(df)
for record in json.loads(df.astype("str").to_json(orient="records")):
print("Added {} records".format(percent_counter)) if (
percent_counter % 50 == 0
) else None
percent_counter += 1
no_approve_record = copy.deepcopy(record)
no_approve_record = {'acc_now_delinq': '1.0', 'acc_open_past_24mths': '2.0', 'annual_inc': '3.0', 'avg_cur_bal': '4.0', 'funded_amnt': '5.0'}
# **note** this is an easy way to interact with a model in a script
response = cdsw.call_model(Model_AccessKey, no_approve_record)
response_labels_sample.append(
{
"uuid": response["response"]["uuid"],
"final_label": label_error(record["label"], percent_counter / percent_max),
"response_label": response["response"]["prediction"],
"timestamp_ms": int(round(time.time() * 1000)),
}
)
#{
# "model_deployment_crn": "crn:cdp:ml:us-west-1:8a1e15cd-04c2-48aa-8f35-b4a8c11997d3:workspace:f5c6e319-47e8-4d61-83bf-2617127acc36/d54e8925-a9e1-4d1f-b7f1-b95961833eb6",
# "prediction": {
# "input_data": "{'acc_now_delinq': '1.0', 'acc_open_past_24mths': '2.0', 'annual_inc': '3.0', 'avg_cur_bal': '4.0', 'funded_amnt': '5.0'}",
# "prediction": "0.0"
# },