Fix typos in tutorials

Author: juliabeliaeva

docs/create_your_first_nn.md

@@ -77,7 +77,7 @@ val model = Sequential.of(
 Quite simple, right? Let's take a closer look at what we have defined here. 
 First, we specify the kind of input we will pass to this neural network. 
 We have images that are 28 x 28 pixels and only have one color channel. 
-Thus the input will be an array of size 28 x 28 x 1.
+Thus, the input will be an array of size 28 x 28 x 1.
 
 The first layer is `Flatten()`. It simply reformats the data, 
 transforming the three-dimensional input array into a one-dimensional array 

docs/importing_keras_model.md

@@ -67,7 +67,7 @@ Non-trainable params: 0
 
 You can find a Jupyter notebook with the Python code used to train this model [here](CIFAR-10.ipynb). 
 
-Once this model has been trained, it's important to save it properly so we can load it from KotlinDL:
+Once this model has been trained, it's important to save it properly, so we can load it from KotlinDL:
 ```python
 model.save('keras-cifar-10/weights', save_format='h5')
 
@@ -124,7 +124,7 @@ fun main() {
 ```
 
 As you can see, loading and using a model trained with Keras is quite simple. 
-If the architecture is supported, all you need to do is save the weights and JSON configuration after training, 
+If the architecture is supported, all you need to do is to save the weights and JSON configuration after training, 
 load it from KotlinDL, and make sure to preprocess your data again before running the inference task.
 
 

docs/loading_trained_model_for_inference.md

@@ -7,7 +7,7 @@ we'll use the test data subset to generate a prediction example that the model h
 The example images in the test data have the same size and format as the ones the model has been trained on, 
 so we do not need to do any additional preprocessing. 
 However, if you are going to train the model on your own data, 
-make sure to use the same image preprocessing before using an image for inference as you did training your model. 
+make sure to use the same image preprocessing before using an image for inference as you did while training your model. 
 Every model expects to get exactly the same input as it was trained on. 
 
 To load the model simply use the path to it, tell it how incoming images should be reshaped (if needed), and call the