|
| ▲ | aesthesia 2 days ago | parent | next [-] |
| The new tokenizer is interesting, but it definitely is possible to adapt a base model to a new tokenizer without too much additional training, especially if you're distilling from a model that uses the new tokenizer. (see, e.g., https://openreview.net/pdf?id=DxKP2E0xK2). |
| |
| ▲ | ACCount37 2 days ago | parent [-] | | Not impossible, but you have to be at least a little bit mad to deploy tokenizer replacement surgery at this scale. They also changed the image encoder, so I'm thinking "new base model". Whatever base that was powering 4.5/4.6 didn't last long then. |
|
|
| ▲ | alecco 2 days ago | parent | prev [-] |
| Yes, I was thinking that. But it could as well be the other way around. Using the pretrained 4.7 (1T?) to speed up ~70% Mythos (10T?) pretraining. It's just speculative decoding but for training. If they did at this scale it's quite an achievement because training is very fragile when doing these kinds of tricks. |
| |
| ▲ | ACCount37 2 days ago | parent [-] | | Reverse distillation. Using small models to bootstrap large models. Get richer signal early in the run when gradients are hectic, get the large model past the early training instability hell. Mad but it does work somewhat. Not really similar to speculative decoding? I don't think that's what they've done here though. It's still black magic, I'm not sure if any lab does it for frontier runs, let alone 10T scale runs. |
|
