if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('--train_path',
type=str,
default='data/cifar10_train.zip',
help='Path to train dataset')
parser.add_argument('--val_path',
type=str,
default='data/cifar10_val.zip',
help='Path to validation dataset')
parser.add_argument('--test_path',
type=str,
default='data/cifar10_test.zip',
help='Path to test dataset')
parser.add_argument(
'--query_path',
type=str,
default='examples/data/image_classification/cifar10_test_1.png',
help='Path(s) to query image(s), delimited by commas')
(args, _) = parser.parse_known_args()
queries = utils.dataset.load_images(args.query_path.split(',')).tolist()
test_model_class(model_file_path=__file__,
model_class='TfVgg16',
task='IMAGE_CLASSIFICATION',
dependencies={ModelDependency.TENSORFLOW: '1.12.0'},
train_dataset_path=args.train_path,
val_dataset_path=args.val_path,
test_dataset_path=args.test_path,
queries=queries)
max_iterations =self._knobs.get('max_iterations')
clf = CRF(
algorithm='lbfgs',
c1=c1,
c2=c2,
max_iterations=max_iterations,
all_possible_transitions=True
)
return clf
if __name__ == '__main__':
test_model_class(
model_file_path=__file__,
model_class='CRFClf',
task='POS_TAGGING',
dependencies={
ModelDependency.SCIKIT_LEARN: '0.20.0',
ModelDependency.NLTK: '3.4.5',
ModelDependency.SKLEARN_CRFSUITE: '0.3.6'
},
train_dataset_path='data/ptb_train.txt',
val_dataset_path='data/ptb_test.txt',
queries=[
['Ms.', 'Haag', 'plays', 'Elianti', '18', '.'],
['The', 'luxury', 'auto', 'maker', 'last', 'year', 'sold', '1,214', 'cars', 'in', 'the', 'U.S.']
]
)
if __name__ == '__main__':
curpath = os.path.join(os.environ['HOME'], 'singa_auto')
os.environ.setdefault('WORKDIR_PATH', curpath)
os.environ.setdefault('PARAMS_DIR_PATH', os.path.join(curpath, 'params'))
train_set_url = os.path.join(curpath, 'data',
'application_train_index.csv')
valid_set_url = train_set_url
test_set_url = os.path.join(curpath, 'data', 'application_test_index.csv')
test_queries = pd.read_csv(test_set_url, index_col=0).iloc[:5]
test_queries = json.loads(test_queries.to_json(orient='records'))
test_model_class(
model_file_path=__file__,
model_class='LightGBM',
task=None,
dependencies={
ModelDependency.TENSORFLOW: '1.12.0',
'lightgbm': '2.3.0',
},
train_dataset_path=train_set_url,
val_dataset_path=valid_set_url,
train_args={
'target': 'TARGET',
'exclude': ['SK_ID_CURR'],
},
queries=test_queries,
)
test_model_class(model_file_path=__file__,
model_class='XgbReg',
task='TABULAR_REGRESSION',
dependencies={ModelDependency.XGBOOST: '0.90'},
train_dataset_path='data/bodyfat_train.csv',
val_dataset_path='data/bodyfat_val.csv',
train_args={
'features': [
'density', 'age', 'weight', 'height', 'neck',
'chest', 'abdomen', 'hip', 'thigh', 'knee',
'ankle', 'biceps', 'forearm', 'wrist'
],
'target':
'bodyfat'
},
queries=[{
'density': 1.0207,
'age': 65,
'weight': 224.5,
'height': 68.25,
'neck': 38.8,
'chest': 119.6,
'abdomen': 118.0,
'hip': 114.3,
'thigh': 61.3,
'knee': 42.1,
'ankle': 23.4,
'biceps': 34.9,
'forearm': 30.1,
'wrist': 19.4
}])
)
return clf
if __name__ == '__main__':
test_model_class(model_file_path=__file__,
model_class='SVCClf',
task='TABULAR_CLASSIFICATION',
dependencies={ModelDependency.SCIKIT_LEARN: '0.20.0'},
train_dataset_path='data/diabetes_train.csv',
val_dataset_path='data/diabetes_val.csv',
train_args={
'features': [
'Pregnancies', 'Glucose', 'BloodPressure',
'SkinThickness', 'Insulin',
'DiabetesPedigreeFunction', 'BMI', 'Age'
],
'target':
'Outcome'
},
queries={
'Pregnancies': 3,
'Glucose': 130,
'BloodPressure': 92,
'SkinThickness': 30,
'Insulin': 90,
'DiabetesPedigreeFunction': 1,
'BMI': 30.4,
'Age': 40
})
-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242],
[-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242]]]])
]
# can not run dist training using this scripts, as each process will be in the container at runtime
test_model_class(model_file_path=__file__,
model_class='DistMinist',
task='IMAGE_CLASSIFICATION',
dependencies={"torch": '1.0.1',
"torchvision": '0.2.2'},
train_dataset_path=args.train_path,
val_dataset_path=args.val_path,
test_dataset_path=args.test_path,
queries=queries,
train_args={"use_dist": False
},
)
"""
Test the model out of singa-auto platform
python -c "import torch;print(torch.cuda.is_available())"
"""
# a = DistMinist()
# model_file = "20.model"
# with open(model_file, 'rb') as f:
# content = f.read()
sc = StandardScaler()
return sc.fit_transform(X)
def _build_classifier(self, n_neighbors, metric, p):
clf = KNeighborsClassifier(n_neighbors=n_neighbors, metric=metric, p=p)
return clf
if __name__ == '__main__':
test_model_class(model_file_path=__file__,
model_class='KNNClf',
task='TABULAR_CLASSIFICATION',
dependencies={ModelDependency.SCIKIT_LEARN: '0.20.0'},
train_dataset_path='data/heart_train.csv',
val_dataset_path='data/heart_test.csv',
queries=[{
'age': 50,
'Sex': '0',
'cp': 3,
'trestbps': 130,
'chol': 220,
'fbs': 1,
'restecg': 0,
'thalach': 170,
'exang': 1,
'oldpeak': 1.7,
'slope': 2,
'ca': 0,
'thal': 3
}])
X = PCA().fit_transform(data)
return X
def _build_classifier(self, penalty, tol, C, fit_intercept, solver):
clf = LogisticRegression(
penalty=penalty,
tol=tol,
C=C,
fit_intercept=fit_intercept,
solver=solver,
)
return clf
if __name__ == '__main__':
test_model_class(model_file_path=__file__,
model_class='LogisticRegClf',
task='TABULAR_CLASSIFICATION',
dependencies={ModelDependency.SCIKIT_LEARN: '0.20.0'},
train_dataset_path='data/diabetes_train.csv',
val_dataset_path='data/diabetes_val.csv',
queries=[{
'Pregnancies': 3,
'Glucose': '130',
'BloodPressure': 92,
'SkinThickness': 30,
'Insulin': 90,
'BMI': 30.4,
'Age': 40
}])
curpath = os.path.join(os.environ['HOME'], 'singa_auto')
os.environ.setdefault('WORKDIR_PATH', curpath)
os.environ.setdefault('PARAMS_DIR_PATH', os.path.join(curpath, 'params'))
train_set_url = os.path.join(curpath, 'data',
'application_train_index.csv')
valid_set_url = train_set_url
test_set_url = os.path.join(curpath, 'data', 'application_test_index.csv')
test_queries = pd.read_csv(test_set_url, index_col=0).iloc[:5]
test_queries = json.loads(test_queries.to_json(orient='records'))
test_model_class(
model_file_path=__file__,
model_class='DNNTorch',
task='TABULAR_CLASSIFICATION',
dependencies={
ModelDependency.TORCH: '1.3.1',
ModelDependency.SCIKIT_LEARN: '0.21.3',
# 'fastai':'0.7.0'
'sklearn-pandas': '1.8.0',
},
train_dataset_path=train_set_url,
val_dataset_path=valid_set_url,
train_args={
'target': 'TARGET',
'exclude': ['SK_ID_CURR'],
},
queries=test_queries,
)
if __name__ == '__main__':
test_model_class(model_file_path=__file__,
model_class='GaussianClf',
task='TABULAR_CLASSIFICATION',
dependencies={ModelDependency.SCIKIT_LEARN: '0.20.0'},
train_dataset_path='data/heart_train.csv',
val_dataset_path='data/heart_val.csv',
train_args={
'features': [
'age', 'sex', 'cp', 'trestbps', 'chol', 'fbs',
'restecg', 'thalach', 'exang', 'oldpeak', 'slope',
'ca', 'thal'],
'target': 'target'
},
queries={
'age': 48,
'sex': 1,
'cp': 2,
'trestbps': 130,
'chol': 225,
'fbs': 1,
'restecg': 1,
'thalach': 172,
'exang': 1,
'oldpeak': 1.7,
'slope': 2,
'ca': 0,
'thal': 3
})
if __name__ == '__main__':
test_model_class(model_file_path=__file__,
model_class='RidgeReg',
task='TABULAR_REGRESSION',
dependencies={ModelDependency.SCIKIT_LEARN: '0.20.0'},
train_dataset_path='data/boston_train.csv',
val_dataset_path='data/boston_val.csv',
train_args={
'features': [
'CRIM', 'ZN', 'INDUS', 'CHAS', 'NOX', 'RM', 'AGE',
'DIS', 'RAD', 'TAX', 'PTRATIO', 'B', 'LSTAT'
],
'target':
'MEDV'
},
queries=[{
'CRIM': 60.1,
'ZN': 0.001,
'INDUS': 18.1,
'CHAS': 0,
'NOX': 597,
'RM': 6.23,
'AGE': 50.0,
'DIS': 1.222,
'RAD': 23,
'TAX': 700,
'PTRATIO': 20.1,
'B': 1.54,
'LSTAT': 11.09
}])
subsample=subsample,
colsample_bytree=colsample_bytree)
else:
clf = xgb.XGBClassifier(n_estimators=n_estimators,
min_child_weight=min_child_weight,
max_depth=max_depth,
gamma=gamma,
subsample=subsample,
colsample_bytree=colsample_bytree,
objective='multi:softmax',
num_class=num_class)
return clf
if __name__ == '__main__':
test_model_class(model_file_path=__file__,
model_class='XgbClf',
task='TABULAR_CLASSIFICATION',
dependencies={ModelDependency.XGBOOST: '0.90'},
train_dataset_path='data/titanic_train.csv',
val_dataset_path='data/titanic_val.csv',
train_args={
'features': ['Pclass', 'Sex', 'Age'],
'target': 'Survived'
},
queries=[{
'Pclass': 1,
'Sex': 'female',
'Age': 2.0
}])
parser = argparse.ArgumentParser()
parser.add_argument('--train_path',
type=str,
default='data/cifar10_train.zip',
help='Path to train dataset')
parser.add_argument('--val_path',
type=str,
default='data/cifar10_val.zip',
help='Path to validation dataset')
parser.add_argument('--test_path',
type=str,
default='data/cifar10_test.zip',
help='Path to test dataset')
parser.add_argument(
'--query_path',
type=str,
default='examples/data/image_classification/0-3096.png',
help='Path(s) to query image(s), delimited by commas')
(args, _) = parser.parse_known_args()
queries = utils.dataset.load_images(args.query_path.split(','))
test_model_class(model_file_path=__file__,
model_class='OnnxResNet18',
task='IMAGE_CLASSIFICATION',
dependencies={ModelDependency.SINGA: '3.0.0',
ModelDependency.ONNX: '1.15.0'},
train_dataset_path=args.train_path,
val_dataset_path=args.val_path,
test_dataset_path=args.test_path,
queries=queries)
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('--train_path',
type=str,
default='data/fashion_mnist_train.zip',
help='Path to train dataset')
parser.add_argument('--val_path',
type=str,
default='data/fashion_mnist_val.zip',
help='Path to validation dataset')
parser.add_argument('--test_path',
type=str,
default='data/fashion_mnist_test.zip',
help='Path to test dataset')
parser.add_argument(
'--query_path',
type=str,
default='examples/data/image_classification/fashion_mnist_test_1.png',
help='Path(s) to query image(s), delimited by commas')
(args, _) = parser.parse_known_args()
queries = utils.dataset.load_images(args.query_path.split(',')).tolist()
test_model_class(model_file_path=__file__,
model_class='SkDt',
task='IMAGE_CLASSIFICATION',
dependencies={ModelDependency.SCIKIT_LEARN: '0.20.0'},
train_dataset_path=args.train_path,
val_dataset_path=args.val_path,
test_dataset_path=args.test_path,
queries=queries)
words_embed_tsr = self._word_embed(words_tsr.view(-1)).view(N, W, Ew)
# Apply dropout to word rep (N x W x Ew)
words_rep_tsr = self._word_dropout(words_embed_tsr)
# Apply bidirectional LSTM to word rep sequence (N x W x 2h)
(words_hidden_rep_tsr, _) = self._word_lstm(words_rep_tsr)
words_hidden_rep_tsr = words_hidden_rep_tsr.contiguous()
# Apply linear + softmax operation for sentence rep for all sentences (N x W x t)
word_probs_tsr = F.softmax(self._word_lin(words_hidden_rep_tsr.view(N * W, self._h * 2)), dim=1).view(N, W, t)
return word_probs_tsr
if __name__ == '__main__':
test_model_class(
model_file_path=__file__,
model_class='PyBiLstm',
task='POS_TAGGING',
dependencies={
ModelDependency.TORCH: '0.4.1'
},
train_dataset_path='data/ptb_train.zip',
val_dataset_path='data/ptb_val.zip',
queries=[
['Ms.', 'Haag', 'plays', 'Elianti', '18', '.'],
['The', 'luxury', 'auto', 'maker', 'last', 'year', 'sold', '1,214', 'cars', 'in', 'the', 'U.S.']
]
)
for col in cat_cols:
df_temp[col] = df[col].map(self._encoding_dict[col])
df = df_temp
return df
def _build_classifier(self, n_estimators, max_depth, oob_score,
max_features):
clf = RandomForestClassifier(n_estimators=n_estimators,
max_depth=max_depth,
oob_score=oob_score,
max_features=max_features)
return clf
if __name__ == '__main__':
test_model_class(model_file_path=__file__,
model_class='RandomForestClf',
task='TABULAR_CLASSIFICATION',
dependencies={ModelDependency.SCIKIT_LEARN: '0.20.0'},
train_dataset_path='data/titanic_train.csv',
val_dataset_path='data/titanic_test.csv',
train_args={
'features': ['Pclass', 'Sex', 'Age'],
'target': 'Survived'
},
queries=[{
'Pclass': 1,
'Sex': 'female',
'Age': 16.0
}])
default='data/val.zip',
help='Path to validation dataset')
parser.add_argument('--test_path',
type=str,
default='data/test.zip',
help='Path to test dataset')
print(os.getcwd())
parser.add_argument(
'--query_path',
type=str,
default=
# 'examples/data/image_classification/xray_1.jpeg,examples/data/image_classification/IM-0103-0001.jpeg,examples/data/image_classification/NORMAL2-IM-0023-0001.jpeg',
# 'examples/data/image_classification/IM-0001-0001.jpeg,examples/data/image_classification/IM-0003-0001.jpeg,examples/data/image_classification/IM-0005-0001.jpeg',
'examples/data/image_classification/cifar10_test_1.png,examples/data/image_classification/cifar10_test_2.png,examples/data/image_classification/fashion_mnist_test_1.png,examples/data/image_classification/fashion_mnist_test_2.png',
help='Path(s) to query image(s), delimited by commas')
(args, _) = parser.parse_known_args()
queries = utils.dataset.load_images(args.query_path.split(',')).tolist()
test_model_class(model_file_path=__file__,
model_class='PyPandaVgg',
task='IMAGE_CLASSIFICATION',
dependencies={
ModelDependency.TORCH: '1.0.1',
ModelDependency.TORCHVISION: '0.2.2'
},
train_dataset_path=args.train_path,
val_dataset_path=args.val_path,
test_dataset_path=args.test_path,
queries=queries)
backpointer = i
# Traverse backpointers to get most probable tags
cur = backpointer
w = len(tokens) - 1
sent_tags = []
while cur is not None:
sent_tags.append(cur)
cur = backpointers[w][cur]
w -= 1
sent_tags.reverse()
sents_tags.append(sent_tags)
return sents_tags
if __name__ == '__main__':
test_model_class(
model_file_path=__file__,
model_class='BigramHmm',
task='POS_TAGGING',
dependencies={},
train_dataset_path='/Users/nailixing/Downloads/data/ptb_train.zip',
val_dataset_path='/Users/nailixing/Downloads/data/ptb_val.zip',
queries=[['Ms.', 'Haag', 'plays', 'Elianti', '18', '.'],
[
'The', 'luxury', 'auto', 'maker', 'last', 'year', 'sold',
'1,214', 'cars', 'in', 'the', 'U.S.'
]])
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('--train_path',
type=str,
default='data/fashion_mnist_train.zip',
help='Path to train dataset')
parser.add_argument('--val_path',
type=str,
default='data/fashion_mnist_val.zip',
help='Path to validation dataset')
parser.add_argument('--test_path',
type=str,
default='data/fashion_mnist_test.zip',
help='Path to test dataset')
parser.add_argument(
'--query_path',
type=str,
default='examples/data/image_classification/fashion_mnist_test_1.png',
help='Path(s) to query image(s), delimited by commas')
(args, _) = parser.parse_known_args()
queries = utils.dataset.load_images(args.query_path.split(',')).tolist()
test_model_class(model_file_path=__file__,
model_class='LeNet5',
task='IMAGE_CLASSIFICATION',
dependencies={ModelDependency.KERAS: '2.2.4'},
train_dataset_path=args.train_path,
val_dataset_path=args.val_path,
test_dataset_path=args.test_path,
queries=queries)
(args, _) = parser.parse_known_args()
# queries = open(args.queries_file_path,'r')
# queries = queries.read().replace("'", "\"")
# queries = json.loads(queries)
# queries = [queries]
queries = [{
'questions': [
'How long individuals are contagious?',
# 'What is the range of the incubation period in humans?',
]
}]
test_model_class(
model_file_path=__file__,
model_class='QuestionAnswering',
task='question_answering_covid19',
# higher version of sentence-transformers and transformers are recommended to avoid 'ndim' and padding issue
dependencies={
ModelDependency.TORCH: '1.0.1',
"torchvision": "0.2.2",
'semanticscholar': '0.1.4',
'sentence_transformers': '0.3.2',
"transformers": '3.0.2',
"tqdm": "4.27",
},
train_dataset_path='',
val_dataset_path='',
fine_tune_dataset_path=args.fine_tune_dataset_path,
queries=queries)
print(args.query_path.split(','))
queries = utils.dataset.load_images(args.query_path.split(','))
test_model_class(
model_file_path=__file__,
model_class='SaMaskRcnn',
task='IMAGE_DETECTION',
dependencies={
"torch": "1.4.0+cu100",
"torchvision": "0.5.0+cu100",
"opencv-python": "4.2.0.34",
"pycocotools": "2.0.1"
},
train_dataset_path=args.train_path,
val_dataset_path=args.val_path,
annotation_dataset_path=args.annotation_dataset_path,
test_dataset_path=None,
train_args={
"num_classes": 3,
"num_epoch": 1,
"dataset_name": "coco2017",
"filter_classes": ["car", 'cat'],
# "dataset_name": "pennfudan",
"batch_size": 2
},
queries=queries)
"""
Test the model out of singa-auto platform
python -c "import torch;print(torch.cuda.is_available())"
"""
