projects
ml
-
safejax
is a Python package to easily serialize JAX (Flax, Haiku, or Objax) model params usingsafetensors
as the tensor-storage format. -
serving-pytorch-models
is a detailed guide on how to set up and deploy TorchServe to expose an inference API for a pre-trained PyTorch CNN. -
serving-tensorflow-models
is a detailed guide on how to set up and deploy TensorFlow Serving to expose an inference API for a pre-trained TensorFlow CNN. -
tensorflow-serving-streamlit
is astreamlit
application to showcase how to deploytensorflow-serving
and send requests to it from a UI. It uses the same model and configuration asserving-tensorflow-models
. -
understanding-resnet
is a PyTorch implementation of ResNet V1 in PyTorch with ported weights fromtimm
and some notes I took for educational/learning purposes. -
ml-monitoring-with-wandb
is a detailed guide on how to train a PyTorch model using PyTorch Lightning and monitoring it using Weights & Biases (wandb).
finance
-
investpy
is a Python package to retrieve data from Investing.com. It used to support all the financial products available at Investing.com, but in late 2022, Investing.com protected their APIs with Cloudflare V2, so sadly this is no longer working… -
investiny
is a simpler, faster, and better-coded version ofinvestpy
, that I developed when the former package was blocked, and I got it working for some weeks, but later it was also blocked. AFAIK it still works if you send a few requests a day, otherwise, you’ll get blacklisted and blocked from sending requests to Investing.com. -
trendet
is a simple Python package to detect trends in time series data. It works withpandas
as well as withinvestpy
as it’s directly integrated with it. Note that now it does not work withinvestpy
anymore due to the blocking from Investing.com, but it does still work with any time series inpandas.DataFrame
.
templates
python-package-template
is a Python package template usingpyproject.toml
,hatch
,pre-commit
,black
,ruff
, andmkdocs
.
others
wandbfsspec
is afsspec
interface for Weights & Biases (wandb) files and artifacts.