Thomas Capelle comments

Results 169 comments of


                                            Thomas Capelle

Organize hyperparameter sweeps

Can I merge this?

Spanish version

isn't block size inferred from the tokenizer ?

Distributed / Parallel version

Not right now, but I would gladly review a PR that adds this feature to the MiniTrainer class. You could use Accelerate or just vanilla DDP.

Undefined symbol

same

feat(peft): add llama token classification example

Hey, can you just use ``` import wandb wandb.login() ``` instead?

feat(peft): add llama token classification example

Two extra small changes: - remove the entity = "saurav" and default to None (so it' uses your default entity after calling `wandb.login`) - Add the image to the peft...

feat(peft): add llama token classification example

extra detail, use a relative path to the image, so when we merge the path to your branch is not needed

feat(peft): add llama token classification example

add it to the folder then

Adds new artifacts colab.

Ping me when ready for final review/merge

Adds new artifacts colab.

> > Why we don't show the classic: > > ```python > > # you can log using the one-liner: > > wandb.log_artifact("file.csv", name="my_artifact", type="data") > > > > #...