ryyzn9
Large Language Models(LLMs)|๐ค|Artificial Intelligence(AI) researcher|Natural Language Processing( NLP )| Computer Vision|Data scientist|ML0ps| python |Pytorch|
Pinned Repositories
accelerate
๐ A simple way to launch, train, and use PyTorch models on almost any device and distributed configuration, automatic mixed precision (including fp8), and easy-to-configure FSDP and DeepSpeed support
AI_to_attend_my_online_classes
So, I made an AI to attend my online classes for me. We all now how boring it gets after a while to attend online classes. So, I made an AI to attend them for me. So, lets get started on what the AI needs to do: These are the things I roughed up the AI needs to do Let's see how will the AI get the data from the class? So from the above image I hope you get the basic gist of how the data collection will work. Lets Start with the data collection and processing: Firstly lets start with the installs we need: Now lets get started with the setup like set up the things we need for the whole code: you will know why we need gender or teacher later in this article Now lets get started with the screenshot part: suppose this is my screen so upon running the code This is the screenshot Now lets take this image and use OpenCV to preprocess it for tesseract OCR: So, basically what the above code does is take the screenshoted image and make it negative so that the black turns white and the white turns black and then detect the the text and draw rectangles around it on the original image. Negative image original image with rectangles Now the image is ready for OCR!! The OCR code is: So, what the above code basically does is use OCR on the preprocessed image and paste the output in a text file hence making a document of whats being said in the class. Here is the recognised file from my AIML class in college: AIML Class 20 Apr 2022.pdf Edit description drive.google.com as you can see the OCR is so fast that it scans the sentences more than once even before they finish (I am in the process of fixing this). now lets make each word a element of a list so that iteration through it is easyer: Now that we have a list lets scan through the words and find if the professor has said my name and respond with yes sir/yes man depending on the gender of the teacher: So, now in the above program the thing thats happening is its going through the list containing all words and searching for my name โEemanโ but as it is captions there might be variation on how the teacher says it so its checking for all the possible spellings that spells Eeman. Now once my name is found it will play a pre-recorded yes sir or yes mam sound according to the gender of the teacher. Now lets get to the notes part: One of the things it wanted this to do is summarise everything that has happened in class at the end of the class. First lets make one using GPT-3: So, in the above code its getting the recognised.txt file from my google drive and turning it into a list and changing it into packets of 4001 tokens and passing it though GPT-3. Then the AI is taking the text and summarising it. Now lets try the GPT-Neo method: So, GPT-3 is paid and me poor and free trial ended so lets make an alternative version in GPT-Neo as it is free ๐ Here is the code its almost same: And with that my notes is part done. How will the AI mute unmute: So, the AI knows when to say Yes sir/Yes mam but it doesn't know how to mute unmute. So the exact co-ordinates of the mute/unmute button in my screen is this: 1223 X 22 The code is pretty simple here it is: Using MacOs so code is so long now my AI can mute/unmute when needed. Now we need my voice because my AI needs to talk in my voice: For that I am going to use Descript. Its an amazing tool used to make a copy of your voice basically and make it say what you want.(Scary but Awesome). Descript creating my voice over dub This process takes time (a lot of time) so you have to wait ๐ let waitโฆ Still waiting ๐ This took more time than I expected. Its done!!! Text can be said in my voice using the app in our case we use the API Once its done I have my voice and use it to make my AI say anything I want! Lets get started with the question detection and answering part: lets start with getting recognised.txt file from my drive: lets download the required libraries: Lets make a function named line-maker to divide the lines as elements of a list: now lets use this function: Now lets google the sentences and find answers: for creative questions lets use GPT-Neo for them: Now our AI can answer questions letsss goooooo. And thats the end of this code. I am going to make a sequel of this which I wanna do stuff with 3d and more believable AI. So, Stay tuned. Here is a recording of my AI attending my class it works well gave me attendance ๐: I f
amber-train
Pre-training code for Amber 7B LLM
annotated_deep_learning_paper_implementations
๐งโ๐ซ 60 Implementations/tutorials of deep learning papers with side-by-side notes ๐; including transformers (original, xl, switch, feedback, vit, ...), optimizers (adam, adabelief, sophia, ...), gans(cyclegan, stylegan2, ...), ๐ฎ reinforcement learning (ppo, dqn), capsnet, distillation, ... ๐ง
arxiv-sanity-lite
arxiv-sanity lite: tag arxiv papers of interest get recommendations of similar papers in a nice UI using SVMs over tfidf feature vectors based on paper abstracts.
DCFormer
deepLense
Deep Regression Techniques for Decoding Dark Matter with Strong Gravitational Lensing Description Strong gravitational lensing is a promising probe of the substructure of dark matter to better understand its underlying nature. Deep learning methods have the potential to accurately identify images containing substructure, and differentiate WIMP part
Galaxy_Zoo-classification-CNN-Transformer
Classify the morphologies of distant galaxies with cnn-vit
halutmatmul_for_windows
Stella Nera is the first Maddness accelerator achieving 15x higher area efficiency (GMAC/s/mm^2) and 25x higher energy efficiency (TMAC/s/W) than direct MatMul accelerators in the same technology
Open-Llama
The complete training code of the open-source high-performance Llama model, including the full process from pre-training to RLHF.
ryyzn9's Repositories
ryyzn9/Galaxy_Zoo-classification-CNN-Transformer
Classify the morphologies of distant galaxies with cnn-vit
ryyzn9/AI_to_attend_my_online_classes
So, I made an AI to attend my online classes for me. We all now how boring it gets after a while to attend online classes. So, I made an AI to attend them for me. So, lets get started on what the AI needs to do: These are the things I roughed up the AI needs to do Let's see how will the AI get the data from the class? So from the above image I hope you get the basic gist of how the data collection will work. Lets Start with the data collection and processing: Firstly lets start with the installs we need: Now lets get started with the setup like set up the things we need for the whole code: you will know why we need gender or teacher later in this article Now lets get started with the screenshot part: suppose this is my screen so upon running the code This is the screenshot Now lets take this image and use OpenCV to preprocess it for tesseract OCR: So, basically what the above code does is take the screenshoted image and make it negative so that the black turns white and the white turns black and then detect the the text and draw rectangles around it on the original image. Negative image original image with rectangles Now the image is ready for OCR!! The OCR code is: So, what the above code basically does is use OCR on the preprocessed image and paste the output in a text file hence making a document of whats being said in the class. Here is the recognised file from my AIML class in college: AIML Class 20 Apr 2022.pdf Edit description drive.google.com as you can see the OCR is so fast that it scans the sentences more than once even before they finish (I am in the process of fixing this). now lets make each word a element of a list so that iteration through it is easyer: Now that we have a list lets scan through the words and find if the professor has said my name and respond with yes sir/yes man depending on the gender of the teacher: So, now in the above program the thing thats happening is its going through the list containing all words and searching for my name โEemanโ but as it is captions there might be variation on how the teacher says it so its checking for all the possible spellings that spells Eeman. Now once my name is found it will play a pre-recorded yes sir or yes mam sound according to the gender of the teacher. Now lets get to the notes part: One of the things it wanted this to do is summarise everything that has happened in class at the end of the class. First lets make one using GPT-3: So, in the above code its getting the recognised.txt file from my google drive and turning it into a list and changing it into packets of 4001 tokens and passing it though GPT-3. Then the AI is taking the text and summarising it. Now lets try the GPT-Neo method: So, GPT-3 is paid and me poor and free trial ended so lets make an alternative version in GPT-Neo as it is free ๐ Here is the code its almost same: And with that my notes is part done. How will the AI mute unmute: So, the AI knows when to say Yes sir/Yes mam but it doesn't know how to mute unmute. So the exact co-ordinates of the mute/unmute button in my screen is this: 1223 X 22 The code is pretty simple here it is: Using MacOs so code is so long now my AI can mute/unmute when needed. Now we need my voice because my AI needs to talk in my voice: For that I am going to use Descript. Its an amazing tool used to make a copy of your voice basically and make it say what you want.(Scary but Awesome). Descript creating my voice over dub This process takes time (a lot of time) so you have to wait ๐ let waitโฆ Still waiting ๐ This took more time than I expected. Its done!!! Text can be said in my voice using the app in our case we use the API Once its done I have my voice and use it to make my AI say anything I want! Lets get started with the question detection and answering part: lets start with getting recognised.txt file from my drive: lets download the required libraries: Lets make a function named line-maker to divide the lines as elements of a list: now lets use this function: Now lets google the sentences and find answers: for creative questions lets use GPT-Neo for them: Now our AI can answer questions letsss goooooo. And thats the end of this code. I am going to make a sequel of this which I wanna do stuff with 3d and more believable AI. So, Stay tuned. Here is a recording of my AI attending my class it works well gave me attendance ๐: I f
ryyzn9/amber-train
Pre-training code for Amber 7B LLM
ryyzn9/annotated_deep_learning_paper_implementations
๐งโ๐ซ 60 Implementations/tutorials of deep learning papers with side-by-side notes ๐; including transformers (original, xl, switch, feedback, vit, ...), optimizers (adam, adabelief, sophia, ...), gans(cyclegan, stylegan2, ...), ๐ฎ reinforcement learning (ppo, dqn), capsnet, distillation, ... ๐ง
ryyzn9/arxiv-sanity-lite
arxiv-sanity lite: tag arxiv papers of interest get recommendations of similar papers in a nice UI using SVMs over tfidf feature vectors based on paper abstracts.
ryyzn9/Autoencoders-AEE
ryyzn9/d2l-en
Interactive deep learning book with multi-framework code, math, and discussions. Adopted at 500 universities from 70 countries including Stanford, MIT, Harvard, and Cambridge.
ryyzn9/deep_lense_for_darkmatter
ryyzn9/deepLense
Deep Regression Techniques for Decoding Dark Matter with Strong Gravitational Lensing Description Strong gravitational lensing is a promising probe of the substructure of dark matter to better understand its underlying nature. Deep learning methods have the potential to accurately identify images containing substructure, and differentiate WIMP part
ryyzn9/Transformers-Tutorials
This repository contains demos I made with the Transformers library by HuggingFace.
ryyzn9/CaImAn
Fully automated pipeline for calcium imaging analysis. (Description updated on Feb.12 2024)
ryyzn9/dbrx
Code examples and resources for DBRX, a large language model developed by Databricks
ryyzn9/Fake-news-detection
ryyzn9/Griffin
Griffin: Mixing Gated Linear Recurrences with Local Attention for Efficient Language Models
ryyzn9/halutmatmul
Hashed Lookup Table based Matrix Multiplication (halutmatmul) - Stella Nera accelerator
ryyzn9/instruct_storyteller_tinyllama2
Training and Fine-tuning an llm in Python and PyTorch.
ryyzn9/JARVIS
JARVIS, a system to connect LLMs with ML community. Paper: https://arxiv.org/pdf/2303.17580.pdf
ryyzn9/llama.cpp
Port of Facebook's LLaMA model in C/C++
ryyzn9/llm-course
Course to get into Large Language Models (LLMs) with roadmaps and Colab notebooks.
ryyzn9/LLMAgentPapers
Must-read Papers on LLM Agents.
ryyzn9/Mathematics-for-Machine-Learning-and-Data-Science-Specialization
Master the Toolkit of AI and Machine Learning. Mathematics for Machine Learning and Data Science is a beginner-friendly Specialization where youโll learn the fundamental mathematics toolkit of machine learning: calculus, linear algebra, statistics, and probability.
ryyzn9/MobiLlama
MobiLlama : Small Language Model tailored for edge devices
ryyzn9/open_llama
OpenLLaMA, a permissively licensed open source reproduction of Meta AIโs LLaMA 7B trained on the RedPajama dataset
ryyzn9/OpenChatKit
ryyzn9/RetNet
An implementation of "Retentive Network: A Successor to Transformer for Large Language Models"
ryyzn9/stanford_alpaca
Code and documentation to train Stanford's Alpaca models, and generate the data.
ryyzn9/start-llms
A complete guide to start and improve your LLM skills in 2024 with little background in the field and stay up-to-date with the latest news and state-of-the-art techniques!
ryyzn9/tensorli
Absolute minimalistic implementation of a GPT-like transformer using only numpy (<650 lines).
ryyzn9/TinyLlama
The TinyLlama project is an open endeavor to pretrain a 1.1B Llama model on 3 trillion tokens.
ryyzn9/yet-another-retnet
A simple but robust PyTorch implementation of RetNet from "Retentive Network: A Successor to Transformer for Large Language Models" (https://arxiv.org/pdf/2307.08621.pdf)