Deep learning Review - MLP/CNN/RNN, self-attention, scaled-dot product attention, multi-head blocks, positional encoding, Transformer architecture, seq2seq task in Colab Text Understanding&Generation in Colab - Encoder-only (BERT) fine-tuning for classification & QA, Decoder-only (GPT-2) text generation, sampling strategies, Seq2seq (T5/BART) summarization & translation, ROUGE Scoring Audio & Music Generation - wav2vec 2.0 for ASR (self-supervised pre-training + fine-tuning), VQ-VAE tokenization for raw audio synthesis, Music Transformer for symbolic melody generation Vision & Video Vision Transformer (ViT) for image classification, CIFAR-10 in Colab; compare to ResNet baseline Latent Diffusion (Stable Diffusion) for text-to-image, Hugging Face Diffusers; generate 256×256 images Video Transformers (ViViT/TimeSformer), Simple Video Transformer for action recognition on short clips.