The deep learning implementation for pytorch starting from basic mnist implementation ranging to chatbots

Hii everyone

I am Aditya,2nd year,IIITA

If you are an AI,ML,DL enthusiast you must be bored of that keras api calling which seems like copying from stackoverflow with understanding background maths. Apart from that there are many issues like implementing advanced NLP and CV architectures are only compatible with tf 1.0 sometimes and also it's doesn't let you feel your worth of understanding maths of AI. if you think calling keras pretrained,preimplemented functions and model will make you AI developer,sry ,that's not the case.

Pytorch is fastest growing, most powerful, handy in both research and development language.Most of recent algorithms are written in it. The most beautiful thing is it reflects the background of models like weights,parameters,gradients that makes you understand what you are doing perfectly.

It is the best language for Deep Learning and yet so simple ( almost like advanced numpy)

I have just prepared a pytorch tutorial for everyone who wanna learn pytorch. I will be uploading three python notebooks with colab link(so that you can play around and learn with it)(ALL on NLP)

First will be introducing pytorch with basic mnist leading to making of Deep leanring's Word2vec(CBOW) and glove word embeddings(all from scratch) and also a sentiment analyzer. I have already done this one and uploaded it.(Don't complain of accuracy etc.. it's based oon very small dataset just for learning purpose)

I will upload few more notebooks, 1st implementing language translater using attention(implementing attention based encoder-decoder from scratch in pytorch),then one implementing Transformer from scratch and the I will try for BERT and GPTs, and the last one , making smart interactive chatbot "FROM ABSOLUTE SCRATCH" in pytorch.

These all are and will be implemented from very scratch and mathematically coded