Course address: course.fast.ai

Lesson 1: Start with image classification
1. Open Jupyter Notebook (Kaggle or Colab recommended)
2. Run the first code example:

```python
from fastai.vision.all import *

# Download dataset
path = untar_data(URLs.PETS)

#Create data loader
dls = ImageDataLoaders.from_name_re(
    path, get_image_files(path/"images"),
    pat=r'([^/]+)_\d+.jpg$',
    item_tfms=Resize(460),
    batch_tfms=aug_transforms(size=224)
)

#Train model
learn = vision_learner(dls, resnet34, metrics=error_rate)
learn.fine_tune(3)
```

3. After a few minutes, the model training is completed
4. Accuracy usually exceeds 90%

Fast.ai NLP Course (Part 3 of course.fast.ai):

1. Introduction to text classification:
```python
from fastai.text.all import *

# Load IMDb data set
path = untar_data(URLs.IMDB)
dls = TextDataLoaders.from_folder(path, valid='test')

# Train language model
learn = text_classifier_learner(
    dls, AWD_LSTM, drop_mult=0.5, metrics=accuracy
)
learn.fine_tune(4, 1e-2)
```

2. Understand the Transformer architecture
3. Implement a simple attention mechanism from scratch

Native PyTorch training loop (~50 lines):


```python


# Need to be written manually:


# - Data loading


# - Model definition


# - loss function


# - Optimizer


# - training loop


# - Validation loop


# - Learning rate scheduling


```




Fastai equivalent code (~5 lines):


```python


dls = ImageDataLoaders.from_folder(path)


learn = vision_learner(dls, resnet50, metrics=accuracy)


learn.lr_find() # Automatically find the optimal learning rate


learn.fine_tune(5)


```