Merge pull request #1068 from pritesh2000/gram-1/04

04_pytorch_custom_datasets.ipynb

Author: mrdbourke

04_pytorch_custom_datasets.ipynb

@@ -166,7 +166,7 @@
    "source": [
     "## 1. Get data\n",
     "\n",
-    "First thing's first we need some data.\n",
+    "First things first we need some data.\n",
     "\n",
     "And like any good cooking show, some data has already been prepared for us.\n",
     "\n",
@@ -270,7 +270,7 @@
     "\n",
     "In our case, we have images of pizza, steak and sushi in standard image classification format.\n",
     "\n",
-    "Image classification format contains separate classes of images in seperate directories titled with a particular class name.\n",
+    "Image classification format contains separate classes of images in separate directories titled with a particular class name.\n",
     "\n",
     "For example, all images of `pizza` are contained in the `pizza/` directory.\n",
     "\n",
@@ -973,7 +973,7 @@
     "\n",
     "We'll do so using [`torch.utils.data.DataLoader`](https://pytorch.org/docs/stable/data.html#torch.utils.data.DataLoader).\n",
     "\n",
-    "Turning our `Dataset`'s into `DataLoader`'s makes them iterable so a model can go through learn the relationships between samples and targets (features and labels).\n",
+    "Turning our `Dataset`'s into `DataLoader`'s makes them iterable so a model can go through and learn the relationships between samples and targets (features and labels).\n",
     "\n",
     "To keep things simple, we'll use a `batch_size=1` and `num_workers=1`.\n",
     "\n",
@@ -1759,7 +1759,7 @@
    "source": [
     "They sure do!\n",
     "\n",
-    "Let's now take a lot at some other forms of data transforms."
+    "Let's now take a look at some other forms of data transforms."
    ]
   },
   {
@@ -1778,7 +1778,7 @@
     "\n",
     "Or cropping it or randomly erasing a portion or randomly rotating them.\n",
     "\n",
-    "Doing this kinds of transforms is often referred to as **data augmentation**.\n",
+    "Doing these kinds of transforms is often referred to as **data augmentation**.\n",
     "\n",
     "**Data augmentation** is the process of altering your data in such a way that you *artificially* increase the diversity of your training set.\n",
     "\n",
@@ -2090,7 +2090,7 @@
     "        self.classifier = nn.Sequential(\n",
     "            nn.Flatten(),\n",
     "            # Where did this in_features shape come from? \n",
-    "            # It's because each layer of our network compresses and changes the shape of our inputs data.\n",
+    "            # It's because each layer of our network compresses and changes the shape of our input data.\n",
     "            nn.Linear(in_features=hidden_units*16*16,\n",
     "                      out_features=output_shape)\n",
     "        )\n",
@@ -2361,7 +2361,7 @@
     "        # 5. Optimizer step\n",
     "        optimizer.step()\n",
     "\n",
-    "        # Calculate and accumulate accuracy metric across all batches\n",
+    "        # Calculate and accumulate accuracy metrics across all batches\n",
     "        y_pred_class = torch.argmax(torch.softmax(y_pred, dim=1), dim=1)\n",
     "        train_acc += (y_pred_class == y).sum().item()/len(y_pred)\n",
     "\n",
@@ -2522,7 +2522,7 @@
     "\n",
     "To keep our experiments quick, we'll train our model for **5 epochs** (though you could increase this if you want).\n",
     "\n",
-    "As for an **optimizer** and **loss function**, we'll use `torch.nn.CrossEntropyLoss()` (since we're working with multi-class classification data) and `torch.optim.Adam()` with a learning rate of `1e-3` respecitvely.\n",
+    "As for an **optimizer** and **loss function**, we'll use `torch.nn.CrossEntropyLoss()` (since we're working with multi-class classification data) and `torch.optim.Adam()` with a learning rate of `1e-3` respectively.\n",
     "\n",
     "To see how long things take, we'll import Python's [`timeit.default_timer()`](https://docs.python.org/3/library/timeit.html#timeit.default_timer) method to calculate the training time."
    ]
@@ -2772,7 +2772,7 @@
    "source": [
     "### 8.1 How to deal with overfitting\n",
     "\n",
-    "Since the main problem with overfitting is that you're model is fitting the training data *too well*, you'll want to use techniques to \"reign it in\".\n",
+    "Since the main problem with overfitting is that your model is fitting the training data *too well*, you'll want to use techniques to \"reign it in\".\n",
     "\n",
     "A common technique of preventing overfitting is known as [**regularization**](https://ml-cheatsheet.readthedocs.io/en/latest/regularization.html).\n",
     "\n",
@@ -2830,7 +2830,7 @@
     "\n",
     "And preventing overfitting and underfitting is possibly the most active area of machine learning research.\n",
     "\n",
-    "Since everone wants their models to fit better (less underfitting) but not so good they don't generalize well and perform in the real world (less overfitting).\n",
+    "Since everyone wants their models to fit better (less underfitting) but not so good they don't generalize well and perform in the real world (less overfitting).\n",
     "\n",
     "There's a fine line between overfitting and underfitting.\n",
     "\n",
@@ -3180,7 +3180,7 @@
     "\n",
     "Even though our models our performing quite poorly, we can still write code to compare them.\n",
     "\n",
-    "Let's first turn our model results in pandas DataFrames."
+    "Let's first turn our model results into pandas DataFrames."
    ]
   },
   {
@@ -3358,7 +3358,7 @@
    "source": [
     "## 11. Make a prediction on a custom image\n",
     "\n",
-    "If you've trained a model on a certain dataset, chances are you'd like to make a prediction on on your own custom data.\n",
+    "If you've trained a model on a certain dataset, chances are you'd like to make a prediction on your own custom data.\n",
     "\n",
     "In our case, since we've trained a model on pizza, steak and sushi images, how could we use our model to make a prediction on one of our own images?\n",
     "\n",