-
- Gradio
-Put a beautiful interactive demo layer over your trained ML model.
-- Get started - See examples - Github -
-
-
-
- Currently, the following
-
Getting started
-- Gradio (gradient input-output) is a python library that allows you to easily transform your trained model into a webpage that easily allows you to interface with it. This allows you to: -
-
-
- Show off the model to others -
- Identify response to unusual behavior -
- Encourage fairness and reproducibility of your model -
Installation
-Gradio is a standard python library, so install it from your terminal like this: -
-
-
-
- pip install gradio
-
- Usage
-To use gradio in your code, simply import it, and then launch the appropriate interface -
-
-
-
- This will open a new browser window which will allow you to draw into a "sketchpad". Behind the scenes, gradio transforms the user drawing into a numeric array and passes it into your model for your prediction.
- Then, gradio will fetch the result and show it to the user. Like this:
-
- import gradio
- import tensorflow as tf
-
- model = tf.keras.Sequential()
- # Define your model, as you ordinarily do
- # Train your model, as you ordinarily do
-
- gradio.interfaces.Sketchpad(model=model, model_type='keras').launch()
-
- - Currently, the following
model_type are supported:
- -
-
- Keras models -
- Sklearn models -
- Arbitrary python functions -
-
Gallery of Interfaces
- -
-
-
-
- 
-
-
-
-
- Draw a Digit: allow users to draw their own digits to test MNIST-type models.
-
-
-
-
-
-
-
-
-
-
- 
-
-
-
-
- Emotion Detector: allow users to make emotions through their webcam and detect their emotion with your model.
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
diff --git a/setup.py b/setup.py
index c646380767..4a9fb8aee2 100644
--- a/setup.py
+++ b/setup.py
@@ -5,7 +5,7 @@ except ImportError:
setup(
name='gradio',
- version='0.9.7',
+ version='0.9.8',
include_package_data=True,
description='Python library for easily interacting with trained machine learning models',
author='Abubakar Abid',
diff --git a/test/test_interface.py b/test/test_interface.py
index 84a19ddd43..fc9afbe4f1 100644
--- a/test/test_interface.py
+++ b/test/test_interface.py
@@ -7,8 +7,7 @@ import gradio.outputs
class TestInterface(unittest.TestCase):
def test_input_output_mapping(self):
- io = gr.Interface(inputs='SketCHPad', outputs='textBOX', fn=lambda
- x: x)
+ io = gr.Interface(inputs='SketCHPad', outputs='textBOX', fn=lambda x: x)
self.assertIsInstance(io.input_interfaces[0], gradio.inputs.Sketchpad)
self.assertIsInstance(io.output_interfaces[0], gradio.outputs.Textbox)
@@ -18,52 +17,15 @@ class TestInterface(unittest.TestCase):
self.assertEqual(io.input_interfaces[0], inp)
def test_output_interface_is_instance(self):
- # out = gradio.outputs.Label(show_confidences=False)
out = gradio.outputs.Label()
io = gr.Interface(inputs='SketCHPad', outputs=out, fn=lambda x: x)
self.assertEqual(io.output_interfaces[0], out)
- def test_keras_model(self):
- try:
- import tensorflow as tf
- except:
- raise unittest.SkipTest("Need tensorflow installed to do keras-based tests")
- inputs = tf.keras.Input(shape=(3,))
- x = tf.keras.layers.Dense(4, activation=tf.nn.relu)(inputs)
- outputs = tf.keras.layers.Dense(5, activation=tf.nn.softmax)(x)
- model = tf.keras.Model(inputs=inputs, outputs=outputs)
- io = gr.Interface(inputs='SketCHPad', outputs='textBOX', fn=model)
- # pred = io.predict(np.ones(shape=(1, 3), ))
- # self.assertEqual(pred.shape, (1, 5))
-
- def test_func_model(self):
+ def test_prediction(self):
def model(x):
return 2*x
- io = gr.Interface(inputs='SketCHPad', outputs='textBOX', fn=model)
- # pred = io.predict(np.ones(shape=(1, 3)))
- # self.assertEqual(pred.shape, (1, 3))
-
- def test_pytorch_model(self):
- try:
- import torch
- except:
- raise unittest.SkipTest("Need torch installed to do pytorch-based tests")
-
- class TwoLayerNet(torch.nn.Module):
- def __init__(self):
- super(TwoLayerNet, self).__init__()
- self.linear1 = torch.nn.Linear(3, 4)
- self.linear2 = torch.nn.Linear(4, 5)
-
- def forward(self, x):
- h_relu = torch.nn.functional.relu(self.linear1(x))
- y_pred = self.linear2(h_relu)
- return y_pred
-
- model = TwoLayerNet()
- io = gr.Interface(inputs='SketCHPad', outputs='textBOX', fn=model)
- # pred = io.predict(np.ones(shape=(1, 3), dtype=np.float32))
- # self.assertEqual(pred.shape, (1, 5))
+ io = gr.Interface(inputs='textbox', outputs='textBOX', fn=model)
+ self.assertEqual(io.predict[0](11), 22)
if __name__ == '__main__':
diff --git a/test/test_outputs.py b/test/test_outputs.py
index a78bcc96d0..f4d854a14b 100644
--- a/test/test_outputs.py
+++ b/test/test_outputs.py
@@ -14,37 +14,39 @@ class TestLabel(unittest.TestCase):
path = outputs.BASE_OUTPUT_INTERFACE_JS_PATH.format(out.__class__.__name__.lower())
self.assertTrue(os.path.exists(os.path.join(PACKAGE_NAME, path)))
- # def test_postprocessing_string(self):
- # string = 'happy'
- # out = outputs.Label()
- # label = json.loads(out.postprocess(string))
- # self.assertDictEqual(label, {outputs.Label.LABEL_KEY: string})
- #
- # def test_postprocessing_1D_array(self):
- # array = np.array([0.1, 0.2, 0, 0.7, 0])
- # true_label = {outputs.Label.LABEL_KEY: 3,
- # outputs.Label.CONFIDENCES_KEY: [
- # {outputs.Label.LABEL_KEY: 3, outputs.Label.CONFIDENCE_KEY: 0.7},
- # {outputs.Label.LABEL_KEY: 1, outputs.Label.CONFIDENCE_KEY: 0.2},
- # {outputs.Label.LABEL_KEY: 0, outputs.Label.CONFIDENCE_KEY: 0.1},
- # ]}
- # out = outputs.Label()
- # label = json.loads(out.postprocess(array))
- # self.assertDictEqual(label, true_label)
+ def test_postprocessing_string(self):
+ string = 'happy'
+ out = outputs.Label()
+ label = out.postprocess(string)
+ self.assertDictEqual(label, {outputs.Label.LABEL_KEY: string})
- # def test_postprocessing_1D_array_no_confidences(self):
- # array = np.array([0.1, 0.2, 0, 0.7, 0])
- # true_label = {outputs.Label.LABEL_KEY: 3}
- # out = outputs.Label(show_confidences=False)
- # label = json.loads(out.postprocess(array))
- # self.assertDictEqual(label, true_label)
- #
- # def test_postprocessing_int(self):
- # true_label_array = np.array([[[3]]])
- # true_label = {outputs.Label.LABEL_KEY: 3}
- # out = outputs.Label()
- # label = json.loads(out.postprocess(true_label_array))
- # self.assertDictEqual(label, true_label)
+ def test_postprocessing_dict(self):
+ orig_label = {
+ 3: 0.7,
+ 1: 0.2,
+ 0: 0.1
+ }
+ true_label = {outputs.Label.LABEL_KEY: 3,
+ outputs.Label.CONFIDENCES_KEY: [
+ {outputs.Label.LABEL_KEY: 3, outputs.Label.CONFIDENCE_KEY: 0.7},
+ {outputs.Label.LABEL_KEY: 1, outputs.Label.CONFIDENCE_KEY: 0.2},
+ {outputs.Label.LABEL_KEY: 0, outputs.Label.CONFIDENCE_KEY: 0.1},
+ ]}
+ out = outputs.Label()
+ label = out.postprocess(orig_label)
+ self.assertDictEqual(label, true_label)
+
+ def test_postprocessing_array(self):
+ array = np.array([0.1, 0.2, 0, 0.7, 0])
+ out = outputs.Label()
+ self.assertRaises(ValueError, out.postprocess, array)
+
+ def test_postprocessing_int(self):
+ label = 3
+ true_label = {outputs.Label.LABEL_KEY: '3'}
+ out = outputs.Label()
+ label = out.postprocess(label)
+ self.assertDictEqual(label, true_label)
class TestTextbox(unittest.TestCase):
diff --git a/upload_to_pypi.sh b/upload_to_pypi.sh
new file mode 100644
index 0000000000..b61a6272c5
--- /dev/null
+++ b/upload_to_pypi.sh
@@ -0,0 +1,10 @@
+#!/bin/bash
+# Before running this, go to setup.py and update the version # of the library
+
+echo "Make sure to update the version number in setup.py!!!"
+rm dist/*
+python setup.py sdist bdist_wheel
+python -m twine upload dist/*
+git add -A
+git commit -m "updated PyPi version"
+git push origin master
-
-
-
-
-
- Automated Author: use an RNN to generate a sequence of words beginning the words that the user has supplied
-
-
-
-
-
-
-
-