Pietro Lesci
pietrolesci
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
updated
a dataset
10 days ago
pietrolesci/pile-deduped
updated
a collection
24 days ago
The Pile Companion
updated
a collection
24 days ago
The Pile Companion
Organizations
Collections
10
spaces
1
models
21
pietrolesci/me100M_finewebedu-20B_bpe32000minipile
Updated
pietrolesci/me100M-tied_finewebedu-20B_bpe32000minipile
Updated
pietrolesci/me850M_minipile_bpe32000minipile
Updated
pietrolesci/me340M-tied_minipile_bpe32000minipile
Updated
pietrolesci/me57M-tied_minipile_bpe2wp32000minipile
Updated
pietrolesci/me57M-tied_minipile_bpe128000minipile
Updated
pietrolesci/me57M-tied_minipile_wordpiece32000minipile
Updated
pietrolesci/me57M-tied_minipile_bpe8064minipile
Updated
pietrolesci/me57M-tied_minipile_bpe32000minipile
Updated
pietrolesci/tokenisers
Updated
datasets
55
pietrolesci/pile-deduped
Viewer
•
Updated
•
748M
•
7.26k
pietrolesci/pythia-deduped-memorisation-profiles
Viewer
•
Updated
•
2.13M
•
61
pietrolesci/pile-validation
Viewer
•
Updated
•
429k
•
214
pietrolesci/pile-deduped-subset
Viewer
•
Updated
•
16.3k
•
33
pietrolesci/pythia-deduped-stats
Viewer
•
Updated
•
16.3M
•
47
pietrolesci/pythia-deduped-stats-raw
Viewer
•
Updated
•
14.9M
•
13.8k
pietrolesci/agnews
Viewer
•
Updated
•
510k
•
86
pietrolesci/amazoncat-13k
Viewer
•
Updated
•
5.99M
•
318
•
1
pietrolesci/wikitoxic
Viewer
•
Updated
•
894k
•
177
•
1
pietrolesci/multiwoz_all_versions
Viewer
•
Updated
•
82k
•
77
•
1