JoaoLages

Maybe you've heard about this technique but you haven't completely understood it, especially the PPO part. This explanation might help.

We will focus on text-to-text language models 📝, such as GPT-3, BLOOM, and T5. Models like BERT, which are encoder-only, are not addressed.

Reinforcement Learning from Human Feedback (RLHF) has been successfully applied in ChatGPT, hence its major increase in popularity. 📈

RLHF is especially useful in two scenarios 🌟:

• You can’t create a good loss function
  • Example: how do you calculate a metric to measure if the model’s output was funny?
• You want to train with production data, but you can’t easily label your production data

MarcoEidinger

Master GitHub markdown tables with code blocks

1. Use HTML tags to define the table to get the best layout result
2. Use either backticks (```) or the HTML pre element with attribute lang
3. Keep a blank line before and after a code block for correct formatting and syntax highlighting

Good

Example: nice looking table to show HTTP Responses

pmbaumgartner

Packing Conda Environments

You must be using conda for this approach. You will need conda installed on the Source machine and the Target machine. The Source machine must have an internet connection, the Target does not. The OS in both environments must match; no going from macOS to Win10 for example.

1. (Source) Install conda-pack in your base python environment.

conda install -c conda-forge conda-pack

kaelzhang

import asyncio

# Ref: https://docs.python.org/3/library/asyncio-eventloop.html#asyncio.get_event_loop
# > If there is no current event loop set in the current OS thread,
# > the OS thread is main, and set_event_loop() has not yet been called,
# > asyncio will create a new event loop and set it as the current one
loop = asyncio.get_event_loop()

# So, loop2 is loop
loop2 = asyncio.get_event_loop()

xeoncross

# Run this in the project repo from the command-line
# http://stackoverflow.com/a/4593065/99923

git log --shortstat --author "Xeoncross" --since "2 weeks ago" --until "1 week ago" | grep "files changed" | awk '{files+=$1; inserted+=$4; deleted+=$6} END {print "files changed", files, "lines inserted:", inserted, "lines deleted:", deleted}'