Why do OpenAI's GPT-2 weights beat mine?

When I finished my project training an LLM from scratch, I was left with a minor mystery. Why were my models worse at instruction-following than the original OpenAI GPT-2 small weights?

I had an evaluation that I was running, based on the instruction fine-tuning code in chapter 7 of "Build a Large Language Model (from Scratch)". The process was to train a model on samples from the Alpaca instruction-following dataset until validation loss started rising, to use that instruction fine-tuned model to generate completions to a held-back test set, and then to use an LLM to compare the results from various different models. The details are here; let's call it the IFT eval.

OpenAI's original weights for GPT-2 small consistently beat my own models, even when mine got better results than theirs on a more technical evaluation, where I just measured the cross entropy loss for each model on a held-back set of test sequences. This surprised me; I would have expected a reasonably close correlation between the two evals -- that better test loss would imply better instruction-following.

I have a couple of thoughts about why this might be, and given that I recently set up poppy, my dedicated LLM training box I decided to inaugurate her with an experiment to test one of them; further experiments will come in time -- though I don't think this will be a focus for the blog. More of a running theme, with occasional posts until either I solve the mystery, or give up in despair...

In this post I'll give a bit more detail about the nature of the problem, and list some of the things I've been thinking might be the cause. In later posts, I'll dig into some of them.

More on the mystery

Let's take a look at the results from my most recent runs of the IFT eval. I've highlighted the OpenAI models in bold, and the table is sorted by the test loss -- that more technical evaluation that I mentioned earlier, where lower is better.

How well the model did with the IFT eval is in the last three…