def test_model_image(sagemaker_session):
knn = KNN(sagemaker_session=sagemaker_session, **ALL_REQ_ARGS)
data = RecordSet("s3://{}/{}".format(BUCKET_NAME, PREFIX), num_records=1, feature_dim=FEATURE_DIM, channel='train')
knn.fit(data, MINI_BATCH_SIZE)
model = knn.create_model()
assert model.image == registry(REGION, "knn") + '/knn:1'
def test_prepare_for_training_wrong_value_lower_mini_batch_size(sagemaker_session):
knn = KNN(base_job_name="knn", sagemaker_session=sagemaker_session, **ALL_REQ_ARGS)
data = RecordSet("s3://{}/{}".format(BUCKET_NAME, PREFIX), num_records=1, feature_dim=FEATURE_DIM,
channel='train')
with pytest.raises(ValueError):
knn._prepare_for_training(data, 0)
def test_predictor_type(sagemaker_session):
knn = KNN(sagemaker_session=sagemaker_session, **ALL_REQ_ARGS)
data = RecordSet("s3://{}/{}".format(BUCKET_NAME, PREFIX), num_records=1, feature_dim=FEATURE_DIM, channel='train')
knn.fit(data, MINI_BATCH_SIZE)
model = knn.create_model()
predictor = model.deploy(1, TRAIN_INSTANCE_TYPE)
assert isinstance(predictor, KNNPredictor)
def test_call_fit_none_mini_batch_size(sagemaker_session):
knn = KNN(base_job_name="knn",
sagemaker_session=sagemaker_session,
**ALL_REQ_ARGS)
data = RecordSet("s3://{}/{}".format(BUCKET_NAME, PREFIX),
num_records=1,
feature_dim=FEATURE_DIM,
channel='train')
knn.fit(data)
def test_model_image(sagemaker_session):
knn = KNN(sagemaker_session=sagemaker_session, **ALL_REQ_ARGS)
data = RecordSet("s3://{}/{}".format(BUCKET_NAME, PREFIX),
num_records=1,
feature_dim=FEATURE_DIM,
channel='train')
knn.fit(data, MINI_BATCH_SIZE)
model = knn.create_model()
assert model.image == registry(REGION, "knn") + '/knn:1'
def test_predictor_type(sagemaker_session):
knn = KNN(sagemaker_session=sagemaker_session, **ALL_REQ_ARGS)
data = RecordSet("s3://{}/{}".format(BUCKET_NAME, PREFIX),
num_records=1,
feature_dim=FEATURE_DIM,
channel='train')
knn.fit(data, MINI_BATCH_SIZE)
model = knn.create_model()
predictor = model.deploy(1, TRAIN_INSTANCE_TYPE)
assert isinstance(predictor, KNNPredictor)
def test_call_fit(base_fit, sagemaker_session):
knn = KNN(base_job_name="knn", sagemaker_session=sagemaker_session, **ALL_REQ_ARGS)
data = RecordSet("s3://{}/{}".format(BUCKET_NAME, PREFIX), num_records=1, feature_dim=FEATURE_DIM, channel='train')
knn.fit(data, MINI_BATCH_SIZE)
base_fit.assert_called_once()
assert len(base_fit.call_args[0]) == 2
assert base_fit.call_args[0][0] == data
assert base_fit.call_args[0][1] == MINI_BATCH_SIZE
def test_prepare_for_training_wrong_value_lower_mini_batch_size(sagemaker_session):
knn = KNN(base_job_name="knn", sagemaker_session=sagemaker_session, **ALL_REQ_ARGS)
data = RecordSet(
"s3://{}/{}".format(BUCKET_NAME, PREFIX),
num_records=1,
feature_dim=FEATURE_DIM,
channel="train",
)
with pytest.raises(ValueError):
knn._prepare_for_training(data, 0)
def test_model_image(sagemaker_session):
knn = KNN(sagemaker_session=sagemaker_session, **ALL_REQ_ARGS)
data = RecordSet(
"s3://{}/{}".format(BUCKET_NAME, PREFIX),
num_records=1,
feature_dim=FEATURE_DIM,
channel="train",
)
knn.fit(data, MINI_BATCH_SIZE)
model = knn.create_model()
assert image_uris.retrieve("knn", REGION) == model.image_uri
def test_all_hyperparameters_regressor(sagemaker_session):
knn = KNN(sagemaker_session=sagemaker_session,
dimension_reduction_type='sign', dimension_reduction_target='2', index_type='faiss.Flat',
index_metric='COSINE', faiss_index_ivf_nlists='auto', faiss_index_pq_m=1, **ALL_REQ_ARGS)
assert knn.hyperparameters() == dict(
k=str(ALL_REQ_ARGS['k']),
sample_size=str(ALL_REQ_ARGS['sample_size']),
predictor_type=str(ALL_REQ_ARGS['predictor_type']),
dimension_reduction_type='sign',
dimension_reduction_target='2',
index_type='faiss.Flat',
index_metric='COSINE',
faiss_index_ivf_nlists='auto',
faiss_index_pq_m='1'
)
def test_init_required_named(sagemaker_session):
knn = KNN(sagemaker_session=sagemaker_session, **ALL_REQ_ARGS)
assert knn.role == COMMON_TRAIN_ARGS["role"]
assert knn.train_instance_count == TRAIN_INSTANCE_COUNT
assert knn.train_instance_type == COMMON_TRAIN_ARGS["train_instance_type"]
assert knn.k == ALL_REQ_ARGS["k"]
def test_call_fit(base_fit, sagemaker_session):
knn = KNN(base_job_name="knn",
sagemaker_session=sagemaker_session,
**ALL_REQ_ARGS)
data = RecordSet("s3://{}/{}".format(BUCKET_NAME, PREFIX),
num_records=1,
feature_dim=FEATURE_DIM,
channel='train')
knn.fit(data, MINI_BATCH_SIZE)
base_fit.assert_called_once()
assert len(base_fit.call_args[0]) == 2
assert base_fit.call_args[0][0] == data
assert base_fit.call_args[0][1] == MINI_BATCH_SIZE
def test_error_optional_hyper_parameters_value(sagemaker_session,
optional_hyper_parameters,
value):
with pytest.raises(ValueError):
test_params = ALL_REQ_ARGS.copy()
test_params.update({optional_hyper_parameters: value})
KNN(sagemaker_session=sagemaker_session, **test_params)
def test_all_hyperparameters_classifier(sagemaker_session):
test_params = ALL_REQ_ARGS.copy()
test_params['predictor_type'] = PREDICTOR_TYPE_CLASSIFIER
knn = KNN(sagemaker_session=sagemaker_session,
dimension_reduction_type='fjlt', dimension_reduction_target='2', index_type='faiss.IVFFlat',
index_metric='L2', faiss_index_ivf_nlists='20', **test_params)
assert knn.hyperparameters() == dict(
k=str(ALL_REQ_ARGS['k']),
sample_size=str(ALL_REQ_ARGS['sample_size']),
predictor_type=str(PREDICTOR_TYPE_CLASSIFIER),
dimension_reduction_type='fjlt',
dimension_reduction_target='2',
index_type='faiss.IVFFlat',
index_metric='L2',
faiss_index_ivf_nlists='20'
)
def test_all_hyperparameters_regressor(sagemaker_session):
knn = KNN(sagemaker_session=sagemaker_session,
dimension_reduction_type='sign',
dimension_reduction_target='2',
index_type='faiss.Flat',
index_metric='COSINE',
faiss_index_ivf_nlists='auto',
faiss_index_pq_m=1,
**ALL_REQ_ARGS)
assert knn.hyperparameters() == dict(
k=str(ALL_REQ_ARGS['k']),
sample_size=str(ALL_REQ_ARGS['sample_size']),
predictor_type=str(ALL_REQ_ARGS['predictor_type']),
dimension_reduction_type='sign',
dimension_reduction_target='2',
index_type='faiss.Flat',
index_metric='COSINE',
faiss_index_ivf_nlists='auto',
faiss_index_pq_m='1')
def test_all_hyperparameters_classifier(sagemaker_session):
test_params = ALL_REQ_ARGS.copy()
test_params['predictor_type'] = PREDICTOR_TYPE_CLASSIFIER
knn = KNN(sagemaker_session=sagemaker_session,
dimension_reduction_type='fjlt',
dimension_reduction_target='2',
index_type='faiss.IVFFlat',
index_metric='L2',
faiss_index_ivf_nlists='20',
**test_params)
assert knn.hyperparameters() == dict(
k=str(ALL_REQ_ARGS['k']),
sample_size=str(ALL_REQ_ARGS['sample_size']),
predictor_type=str(PREDICTOR_TYPE_CLASSIFIER),
dimension_reduction_type='fjlt',
dimension_reduction_target='2',
index_type='faiss.IVFFlat',
index_metric='L2',
faiss_index_ivf_nlists='20')
def test_all_hyperparameters_regressor(sagemaker_session):
knn = KNN(sagemaker_session=sagemaker_session,
dimension_reduction_type="sign",
dimension_reduction_target="2",
index_type="faiss.Flat",
index_metric="COSINE",
faiss_index_ivf_nlists="auto",
faiss_index_pq_m=1,
**ALL_REQ_ARGS)
assert knn.hyperparameters() == dict(
k=str(ALL_REQ_ARGS["k"]),
sample_size=str(ALL_REQ_ARGS["sample_size"]),
predictor_type=str(ALL_REQ_ARGS["predictor_type"]),
dimension_reduction_type="sign",
dimension_reduction_target="2",
index_type="faiss.Flat",
index_metric="COSINE",
faiss_index_ivf_nlists="auto",
faiss_index_pq_m="1",
)
def test_all_hyperparameters_classifier(sagemaker_session):
test_params = ALL_REQ_ARGS.copy()
test_params["predictor_type"] = PREDICTOR_TYPE_CLASSIFIER
knn = KNN(sagemaker_session=sagemaker_session,
dimension_reduction_type="fjlt",
dimension_reduction_target="2",
index_type="faiss.IVFFlat",
index_metric="L2",
faiss_index_ivf_nlists="20",
**test_params)
assert knn.hyperparameters() == dict(
k=str(ALL_REQ_ARGS["k"]),
sample_size=str(ALL_REQ_ARGS["sample_size"]),
predictor_type=str(PREDICTOR_TYPE_CLASSIFIER),
dimension_reduction_type="fjlt",
dimension_reduction_target="2",
index_type="faiss.IVFFlat",
index_metric="L2",
faiss_index_ivf_nlists="20",
)
def test_init_required_positional(sagemaker_session):
knn = KNN(ROLE,
TRAIN_INSTANCE_COUNT,
TRAIN_INSTANCE_TYPE,
K,
SAMPLE_SIZE,
PREDICTOR_TYPE_REGRESSOR,
sagemaker_session=sagemaker_session)
assert knn.role == ROLE
assert knn.train_instance_count == TRAIN_INSTANCE_COUNT
assert knn.train_instance_type == TRAIN_INSTANCE_TYPE
assert knn.k == K
def test_predictor_custom_serialization(sagemaker_session):
knn = KNN(sagemaker_session=sagemaker_session, **ALL_REQ_ARGS)
data = RecordSet(
"s3://{}/{}".format(BUCKET_NAME, PREFIX),
num_records=1,
feature_dim=FEATURE_DIM,
channel="train",
)
knn.fit(data, MINI_BATCH_SIZE)
model = knn.create_model()
custom_serializer = Mock()
custom_deserializer = Mock()
predictor = model.deploy(
1,
INSTANCE_TYPE,
serializer=custom_serializer,
deserializer=custom_deserializer,
)
assert isinstance(predictor, KNNPredictor)
assert predictor.serializer is custom_serializer
assert predictor.deserializer is custom_deserializer
def test_error_conditional_hyper_parameters_value(
sagemaker_session, conditional_hyper_parameters):
with pytest.raises(ValueError):
test_params = ALL_REQ_ARGS.copy()
test_params.update(conditional_hyper_parameters)
KNN(sagemaker_session=sagemaker_session, **test_params)
def test_required_hyper_parameters_value(sagemaker_session,
required_hyper_parameters, value):
with pytest.raises(ValueError):
test_params = ALL_REQ_ARGS.copy()
test_params[required_hyper_parameters] = value
KNN(sagemaker_session=sagemaker_session, **test_params)
def test_image(sagemaker_session):
knn = KNN(sagemaker_session=sagemaker_session, **ALL_REQ_ARGS)
assert knn.train_image() == registry(REGION, "knn") + "/knn:1"
def test_call_fit_none_mini_batch_size(sagemaker_session):
knn = KNN(base_job_name="knn", sagemaker_session=sagemaker_session, **ALL_REQ_ARGS)
data = RecordSet("s3://{}/{}".format(BUCKET_NAME, PREFIX), num_records=1, feature_dim=FEATURE_DIM,
channel='train')
knn.fit(data)
def test_image(sagemaker_session):
knn = KNN(sagemaker_session=sagemaker_session, **ALL_REQ_ARGS)
assert image_uris.retrieve("knn", REGION) == knn.training_image_uri()
def test_image(sagemaker_session):
knn = KNN(sagemaker_session=sagemaker_session, **ALL_REQ_ARGS)
assert knn.train_image() == registry(REGION, "knn") + '/knn:1'
