Someone in the EleutherAI discord is reporting that finetuning Pythia models of different sizes on the same data in the same order is giving spookily similar loss curves, just vertically shifted.
[image]Will training Pythia models from scratch in the same way produce similar behaviour? Resolves N/A if it turns out the original result was just a bug or something like that.
| Indicator | Value |
|---|---|
| Stars | ★★☆☆☆ |
| Platform | Manifold Markets |
| Forecasters | 13 |
| Volume | M803 |
Someone in the EleutherAI discord is reporting that finetuning Pythia models of different sizes on the same data in the same order is giving spookily similar loss curves, just vertically shifted.
[image]Will training Pythia models from scratch in...