This section covers the integration of the mnist dataset into Tensorleap. We'll later use this dataset with a classification model.
Dataset Script
Below is the full dataset script to be used in the integration. More information about the structure of this script can be found under Dataset Script.
from typing import List
import numpy as np
from sklearn.model_selection import train_test_split
from tensorflow.keras.datasets import mnist
from tensorflow.keras.utils import to_categorical
# Tensorleap imports
from code_loader import leap_binder
from code_loader.contract.datasetclasses import PreprocessResponse
from code_loader.contract.enums import Metric, DatasetMetadataType
# Preprocess Function
def preprocess_func() -> List[PreprocessResponse]:
(data_X, data_Y), (test_X, test_Y) = mnist.load_data()
data_X = np.expand_dims(data_X, axis=-1) # Reshape :,28,28 -> :,28,28,1
data_X = data_X / 255 # Normalize to [0,1]
data_Y = to_categorical(data_Y) # Hot Vector
test_X = np.expand_dims(test_X, axis=-1) # Reshape :,28,28 -> :,28,28,1
test_X = test_X / 255 # Normalize to [0,1]
test_Y = to_categorical(test_Y) # Hot Vector
train_X, val_X, train_Y, val_Y = train_test_split(data_X, data_Y, test_size=0.2, random_state=42)
# Generate a PreprocessResponse for each data slice, to later be read by the encoders.
# The length of each data slice is provided, along with the data dictionary.
# In this example we pass `images` and `labels` that later are encoded into the inputs and outputs
train = PreprocessResponse(length=len(train_X), data={'images': train_X, 'labels': train_Y})
val = PreprocessResponse(length=len(val_X), data={'images': val_X, 'labels': val_Y})
test = PreprocessResponse(length=len(test_X), data={'images': test_X, 'labels': test_Y})
response = [train, val, test]
return response
# Input encoder fetches the image with the index `idx` from the `images` array set in
# the PreprocessResponse data. Returns a numpy array containing the sample's image.
def input_encoder(idx: int, preprocess: PreprocessResponse) -> np.ndarray:
return preprocess.data['images'][idx].astype('float32')
# Ground truth encoder fetches the label with the index `idx` from the `labels` array set in
# the PreprocessResponse's data. Returns a numpy array containing a hot vector label correlated with the sample.
def gt_encoder(idx: int, preprocess: PreprocessResponse) -> np.ndarray:
return preprocess.data['labels'][idx].astype('float32')
# Metadata functions allow to add extra data for a later use in analysis.
# This metadata adds the int digit of each sample (not a hot vector).
def metadata_label(idx: int, preprocess: PreprocessResponse) -> int:
one_hot_digit = gt_encoder(idx, preprocess)
digit = one_hot_digit.argmax()
digit_int = int(digit)
return digit_int
LABELS = ['0','1','2','3','4','5','6','7','8','9']
# Dataset binding functions to bind the functions above to the `Dataset Instance`.
leap_binder.set_preprocess(function=preprocess_func)
leap_binder.set_input(function=input_encoder, name='image')
leap_binder.set_ground_truth(function=gt_encoder, name='classes')
leap_binder.set_metadata(function=metadata_label, metadata_type=DatasetMetadataType.int, name='label')
leap_binder.add_prediction(name='classes', labels=LABELS)
Script: copy and paste the script from the Dataset Script above
Click Save.
After saving the mnist dataset, the platform will automatically parse the dataset script. This process evaluates the script and ensures that all its functions, including the ability to successfully read the data, are working as expected.
Upon successful parsing, the details of the MNIST dataset will be displayed on the right. In case of unsuccessful parsing, errors will be shown instead.
Initial CLI Setup
Verify that leapcli is installed. For more information, see Installing Leap CLI.
Project Folder Setup
Create a folder for our mnist project.
mkdir mnist
cd mnist
Initialize and synchronize the created folder with the Tensorleap platform by running a command that will set up the .tensorleap folder within the project folder. The command leap init (PROJECT) (DATASET) (--h5/--onnx) with the following parameters:
PROJECT = MNIST (project name)
DATASET = mnist (dataset name)
(--h5/--onnx) = model format, --h5 for Tensorflow (H5) and --onnx for PyTorch (ONNX)
leap init MNIST mnist --h5
Next, we need to set your credentials to leap CLI by running the following command:
leap login [API_ID] [API_KEY] [ORIGIN]
Push Dataset
When using the CLI, the Dataset Script is defined within the .tensorleap/dataset.py file, and the Dataset Instance is created/updated upon performing leap push.
By default, the .tensorleap/dataset.py file has a sample template. Let's replace it with our Dataset Script above. One way to do it is with vim:
Next, we'll push our dataset to the Tensorleap platform using the following command:
leap push --dataset
It should print out:
New dataset detected. Dataset name: mnist
Push command successfully complete
Congrats! You have successfuly created the mnistDataset Instance and integrated the Dataset Script. You can view it in the UI in the Resources Management view.
Up Next - Model Integration
The purpose of this section was to help you define a dataset script and create a dataset instance in Tensorleap.
Now that the mnist dataset has been integrated into Tensorleap, we can use it with a classification model. That's what we'll do in the next section, where we'll build a classification model.
