undefined | Better HN

0 pointspaulhilbert1y ago0 comments

No, I would argue that from the three main ingredients - training data, model source code and weights - weights are the furthest away from something akin to source code.

They're more like obfuscated binaries. When it comes to fine-tuning only however things shift a little bit, yes.

0 comments

sidcool1y ago

I don't expect them to release the data used to train the models. But I agree that the code is an important ingredient of 'open'.

frabcus1y ago

Must include the code that curates the data

j / k navigate · click thread line to collapse

0 comments

sidcool1y ago

I don't expect them to release the data used to train the models. But I agree that the code is an important ingredient of 'open'.

frabcus1y ago

Must include the code that curates the data

j / k navigate · click thread line to collapse