dnn-tech-DS 0.1.0

A grab-bag of ready-to-copy example scripts for common deep-learning architectures (NN, CNN, ResNet, RNN/LSTM, GRU, AutoEncoder, VAE, text classification, GAN, DQN).

dnn-tech

A grab-bag of ready-to-copy example scripts covering common deep-learning architectures: basic feed-forward NN, CNN, ResNet, RNN/LSTM, GRU sentiment model, AutoEncoder, VAE image generation, text classification, GAN, and DQN reinforcement learning — all on MNIST/toy data so they run out of the box.

Each example is stored as a string and printed to stderr when you call the corresponding function, so you can quickly copy a working starting point for a given architecture.

Install

pip install dnn-tech

(For local development, from the repo root:)

pip install -e .

Usage

import dnn_tech

dnn_tech.commands()      # list every available example
dnn_tech.p2_CNN()        # print the CNN example script (goes to stderr)
dnn_tech.all_lib()       # print `pip install` commands for the libraries
                          # used across the examples (tensorflow, torch, ...)

Or from the command line:

dnn-tech                 # list available example functions
dnn-tech p2_CNN           # print the CNN example script
dnn-tech all_lib           # print pip install commands for dependencies

Available examples

Function	Architecture
p1_BasicNN()	Basic feed-forward NN (Keras, MNIST)
p2_CNN()	CNN (Keras, MNIST)
p3_Resnet()	ResNet-50 (PyTorch, torchvision)
p4_RNN_LSTM()	RNN/LSTM (Keras, toy sequence data)
p42_RNN_LSTM()	RNN/LSTM (Keras, sine-wave forecasting)
p5_LSTM()	Char-level LSTM text generation
p6_GRU_Sentiment()	GRU sentiment classifier
p7_AutoEncoder()	AutoEncoder
p8_vae_imggen()	Variational AutoEncoder image gen
p10_text_classification()	Text classification
p11_gan()	GAN
p12_deepq()	Deep Q-Network (PyTorch, CartPole)
all_lib()	Print pip install commands
commands()	List all available commands

Note: commands() mentions a p9_trans() (Transformer example), but that function isn't implemented in the source yet — calling it will raise AttributeError. Add it to src/dnn_tech/core.py and re-export it from __init__.py when it's ready.

License

MIT