This is a small NumPy-based deep learning library implementation, mimicing PyTorch. It originates from the course 11-785 Introduction to Deep Learning at Carnegie Mellon Fall 2021.
- Activation: (identity, sigmoid, tanh, ReLU)
- Batchnorm
- SoftmaxCrossEntropy
- Linear Layer
- Flatten Layer
- Multi-layer Perceptron
- Convolution Layer: (1D, 2D, with/without dilation and upsampling)
- CNN+MLP model
- Recurrent Unit Cell
- Gate Recurrent Unit Cell
- Search: (greedy Search, beam Search)
- Connectionist Temporal Classification Cell (CTC)
- CTC Loss
- Locked Dropout Layer
- Embedding Dropout Layer
- Language Model (prediction, generation)