yoavg

Reinforcement Learning for Language Models

Yoav Goldberg, April 2023.

Why RL?

With the release of the ChatGPT model and followup large language models (LLMs), there was a lot of discussion of the importance of "RLHF training", that is, "reinforcement learning from human feedback". I was puzzled for a while as to why RL (Reinforcement Learning) is better than learning from demonstrations (a.k.a supervised learning) for training language models. Shouldn't learning from demonstrations (or, in language model terminology "instruction fine tuning", learning to immitate human written answers) be sufficient? I came up with a theoretical argument that was somewhat convincing. But I came to realize there is an additional argumment which not only supports the case of RL training, but also requires it, in particular for models like ChatGPT. This additional argument is spelled out in (the first half of) a talk by John Schulman from OpenAI. This post pretty much

Jonalogy

Handling Multiple Github Accounts on MacOS

The only way I've succeeded so far is to employ SSH.

Assuming you are new to this like me, first I'd like to share with you that your Mac has a SSH config file in a .ssh directory. The config file is where you draw relations of your SSH keys to each GitHub (or Bitbucket) account, and all your SSH keys generated are saved into .ssh directory by default. You can navigate to it by running cd ~/.ssh within your terminal, open the config file with any editor, and it should look something like this:

Host *
 AddKeysToAgent yes

> UseKeyChain yes