In this repository I explore and visualize how ML mechanisms (algoritms, layers, metrics, etc.) work.
- 📩Convolution Layers.
What does image look like when it passes through convolution and pooling layer? How different are activations of different feature maps? - 📅Learning Rate Schedulers.
How learning rate changes at every step thanks with different schedulers? (Now only the scheduler from LLama2 and ReduceLROnPlateau). - ⚠Attention.
Are results of Query*Key multiplication really higher for words that are related? Heatmapping scores for Self-Attention and Encoder-Decoder Attention of T5. - 🤘ROC curves.
What do ROC curves look like for differrnt (true, pred) combinations? How do they differ from other metrics?