arrmansa,github

auto-function-serving

A python package to offload a function call to an http server running on localhost automatically using a decorator. Compatible with multiprocessing, pickle, flask, fastapi, async etc..

basic-ui-for-gpt-j-6b-with-low-vram

A repository to run gpt-j-6b on low vram machines (4.2 gb minimum vram for 2000 token context, 3.5 gb for 1000 token context). Model loading takes 12gb free ram.

basic-ui-for-gpt-neo-with-low-vram

A basic ui for running gpt neo 2.7B on low vram (3 gb Vram minimum)

calculating-pi-by-counting-collisions-with-numba

Based on a 3Blue1Brown video and G Galperin's work. 150x speedup with numba.

chain-reaction-python

Some tests on the game of chain reaction

codenames.game-response-decode

Decodes Network responses of https://codenames.game/ and reveals answers.

exponential-distribution-interval

using dice rolls, chunked dice rolls and the normal method

fast-fizzbuzz-python

FizzBuzz implementation in python that outputs to pv > /dev/null.

fast-gematria-calculator

tests on https://codegolf.stackexchange.com/questions/252679/fastest-gematria-calculator

fast-sentence-transformers

This repository, called fast sentence transformers, contains code to run 5X faster sentence transformers using tools like quantization and ONNX.

fast-word-equation-solver

https://codegolf.stackexchange.com/questions/34926/write-a-fast-word-equation-solver

fastapi-logging-tests

log unique request id, while replacing newlines for a good log format

godai-server-limited-vram

Limited Vram mod for GodAI currently not working.

gpt-neo-limited-vram-cuda

A notebook that runs GPT-Neo with low vram (6 gb) and cuda acceleration by loading it into gpu memory in smaller parts.

gpt-neo_dungeon

Colab notebooks to run a basic AI Dungeon clone using gpt-neo-2.7B

gpu-bandwidth-benchmark

A test to see the speed of transfer of tensors from cpu to gpu in pytorch with 2 cuda streams

leetcode-nice-solutions

Some solutions for leetcode problems that I found fun - either the algorithm or using numpy bruteforce where it shouldn't be

metaballs

Just some tests after watching https://www.youtube.com/watch?v=6oMZb3yP_H8 Cool stuff.

A python/flask/html/js/css frontend to huggingface's Transformers, allowing a model (e.g.: GPT2/Neo) to be loaded for interactive inference. Heavily under construction. Intending to showcase new or interesting ways of generating texts.

model-compression-by-partially-connected-layers

Decrease the model size by replacing fully connected layers with partially connected layers

modify-tuples-strings-inplace-python

Methods to modify python immutable python objects inplace using ctypes. Experimental. Warning - Do not use these in your code unless you are very, very sure what you are doing and have very good reasons.