def test_predict():
test_review = 'The simplest pleasures in life are the best, and this film is one of them. Combining a rather basic storyline of love and adventure this movie transcends the usual weekend fair with wit and unmitigated charm.'
model_dir = os.path.join(here, './data/modelDir')
model = model_fn(model_dir)
print(model)
data_X, data_len = convert_and_pad(model.word_dict, test_review)
print('data_X ')
print(data_X)
print('data_len')
print(data_len)
# Using data_X and data_len we construct an appropriate input tensor. Remember
# that our model expects input data of the form 'len, review[500]'.
data_pack = np.hstack((data_len, data_X))
print('data_pack (shape: {})'.format(data_pack.shape))
print(data_pack)
data_pack = data_pack.reshape(1, -1)
print('data_pack reshaped (shape: {})'.format(data_pack.shape))
print(data_pack)
data = torch.from_numpy(data_pack)
print('data (shape: {})'.format(data.shape))
print(data)
input_data = 'The simplest pleasures in life are the best, and this film is one of them. Combining a rather basic storyline of love and adventure this movie transcends the usual weekend fair with wit and unmitigated charm.'
result = predict_fn(input_data, model)
print(result)
def test_serve():
model_dir = os.path.join(here, './data/modelDir')
model = model_fn(model_dir)
print(model)
# input = input_fn(pickle.dumps(u"Best movie ever"), 'text/plain')
# print(input)
input_data = "Best movie ever"
predict_fn(input_data=input_data, model=model)
parser.add_argument('--img-width', type=int, default=320, metavar='N',
help='width of image (default: 128)')
parser.add_argument('--img-height', type=int, default=180, metavar='N',
help='height of image (default: 72)')
parser.add_argument('--batch-size', type=int, default=8, metavar='N',
help='input batch size for training (default: 8)')
args = parser.parse_args()
print(f'Data Dir: {args.data_dir}')
print(f'Model Dir: {args.model_dir}')
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
use_gpu = torch.cuda.is_available()
model = model_fn(args.model_dir)
model_info = get_model_info(args.model_dir)
args = parser.parse_args()
# remove last fully-connected layer to get feaut
new_classifier = nn.Sequential(*list(model.classifier.children())[:-1])
model.classifier = new_classifier
model.train(False)
model.eval()
# retrieves data loaders and datasets, and the label names
dataloader, dataset_size = get_data_loaders(img_dir=args.data_dir, img_height=args.img_height, img_width=args.img_width, batch_size=args.batch_size)
X, y = get_feature_matrix_from_dataset(dataloader)
""" Call Predict function on a loaded LSTM model"""
if request.method == "POST":
LOG.info("I am a post")
if request.form:
LOG.info("I have form data")
#print(request.form['kommentar'])
if request.data:
LOG.info("I have data")
LOG.info(request.data)
if request.json:
LOG.info("I have json")
# Do stuff with the data...
return jsonify({"message": "OK"})
else:
LOG.info("fail")
data = request.data
LOG.info("Form data is: \n %s" % data.decode('utf-8'))
# get an output prediction from the pretrained model, model
result = predict_fn(data.decode('utf-8'), model)
LOG.info("Prediction value is: %s" % result)
return str(result)
if __name__ == "__main__":
# load pretrained model as model
model = model_fn("./model")
app.run(host='0.0.0.0', port=80, debug=True) # specify port=80