Matt

Hm, those are some fairly large discrepancies! The debugging process we recommend when something like that happens is: - Make a test environment and load the PT and TF models...

Yep, those flax issues are unrelated, just ignore them. I'll review everything today, but the CI looks good!

Just looked over the last few changes - I'm happy to merge it at this point. Thanks again for putting in the work on this!

Hi @mollerup23, mostly looks good! One thing to watch out for is that in some cases the default value of the argument has been changed. It's easy to see if...

Extremely late reply on the TF front, but yeah, we aren't really running TF models in half precision right now. We do support mixed precision (similar to Torch AMP), but...

Hey @amyeroberts, is this PR still going ahead? It looked almost ready!

@amyeroberts Ah, that makes sense! It's totally okay to upload your own super-mini model and use that - it doesn't really matter if the accuracy is bad, the test will...

Hey @daniel-furman that table is great! Can I use it in a tweet about chat templates?

That's surprising! I would expect that chat models should work much better in basically all cases when their correct template is used

Hi @code-isnot-cold, great question! The short answer is that the text generation pipeline will only generate one sample at a time, so you won't gain any benefit from batching samples...