metadata

metadata copied to clipboard

shanyas10

Evaluation toxicity for website description and data source

1

When finished share with Christopher and Shanya.

norakassner

simple zero-shot eval function: website description

1

ppl on website specific testset. Contact @cccntu and Christopher

norakassner

Evaluation bias

1

norakassner

simple zero-shot eval function: generation length

1

norakassner

perf: collator with padding for tpu

9

Imitating https://huggingface.co/docs/accelerate/quicktour.html#training-on-tpu I also see that there's also a padding procedure in https://github.com/bigscience-workshop/metadata/blob/master/bsmetadata/experiments/sample.py#L16 Not really sure: - what we want here for both TPU and GPU (which can be longest...

tianjianjiang

norakassner

When done e-mail @VictorSanh Find temaplte here: https://github.com/bigscience-workshop/metadata/tree/master/experiments/jz/templates/SLURM

norakassner

While testing the real data extraction, I encountered a new problem: the websites descriptions are rarely present in the `metadata_website_desc` column. Therefore, the datasets library cannot load such a dataset...

SaulLu

norakassner