Pietro Lesci's picture

15 4 9

Pietro Lesci

pietrolesci

·

https://pietrolesci.github.io/

AI & ML interests

I like developing and applying causal methods to study the effect of training choices on models’ behaviour, including memorisation, shortcut learning, and tokenisation.

Recent Activity

upvoted an article about 2 months ago

There is no such thing as a tokenizer-free lunch

updated a model 2 months ago

pietrolesci/tokenisers

updated a model 2 months ago

pietrolesci/tokenizers

View all activity

Organizations

pietrolesci 's datasets 56

pietrolesci/unimixlm

Viewer • Updated Jul 25 • 81.9M • 463

pietrolesci/me-minipile-evals

Viewer • Updated Jun 3 • 1.22M • 15

pietrolesci/pile-deduped

Viewer • Updated May 5 • 748M • 35

pietrolesci/pythia-deduped-memorisation-profiles

Viewer • Updated Apr 9 • 2.13M • 44

pietrolesci/pile-validation

Viewer • Updated Apr 9 • 429k • 21

pietrolesci/pile-deduped-subset

Viewer • Updated Apr 9 • 16.3k • 45

pietrolesci/pythia-deduped-stats

Viewer • Updated Apr 9 • 16.3M • 153

pietrolesci/pythia-deduped-stats-raw

Viewer • Updated Apr 9 • 14.9M • 29.6k

pietrolesci/agnews

Viewer • Updated Apr 9 • 510k • 23

pietrolesci/amazoncat-13k

Viewer • Updated Apr 9 • 5.99M • 2.48k • 1

pietrolesci/wikitoxic

Viewer • Updated Apr 9 • 894k • 31 • 1

pietrolesci/multiwoz_all_versions

Viewer • Updated Apr 9 • 82k • 41 • 1

pietrolesci/anchoral-paper-artefacts

Viewer • Updated Apr 9 • 2.78M • 362

pietrolesci/pile-deduped-pythia-preshuffled

Viewer • Updated Mar 25 • 244M • 803

pietrolesci/pile-deduped-pythia-tokfreq

Viewer • Updated Mar 17 • 50.1k • 8

pietrolesci/finewebedu-20B

Viewer • Updated Mar 16 • 40.4M • 1.67k

pietrolesci/minipile

Viewer • Updated Feb 27 • 6.06M • 155

pietrolesci/opus-5langs-1M

Viewer • Updated Dec 10, 2024 • 5M • 22

pietrolesci/opus-raw

Viewer • Updated Nov 27, 2024 • 4.06B • 1.87k

pietrolesci/pythia-pile-stats

Viewer • Updated Sep 23, 2024 • 113M • 11

pietrolesci/slim-pajama-eval

Viewer • Updated Sep 16, 2024 • 1.84M • 14 • 1

pietrolesci/pile-subset

Updated Sep 13, 2024 • 37

pietrolesci/cmnist

Viewer • Updated Jul 29, 2024 • 308k • 15

pietrolesci/celeba-wilds

Viewer • Updated Jul 2, 2024 • 203k • 25 • 1

pietrolesci/civilcomments-wilds

Viewer • Updated Jul 2, 2024 • 893k • 41 • 2

pietrolesci/mnli-stats

Viewer • Updated May 13, 2024 • 785k • 13

pietrolesci/mnli-embeddings

Viewer • Updated Mar 22, 2024 • 785k • 8

pietrolesci/_mnli-stats

Viewer • Updated Mar 20, 2024 • 15.7M • 7

pietrolesci/wikitext-103-raw-v1_gpt2-20k

Viewer • Updated Nov 16, 2023 • 8.01M • 25

pietrolesci/yahoo_answers_topics

Viewer • Updated Sep 25, 2023 • 2.92M • 20