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 model
about 5 hours ago
pietrolesci/small_unigramlm128k
updated
a collection
about 5 hours ago
UnimixLM
published
a model
about 5 hours ago
pietrolesci/small_unigramlm128k
Organizations
models
25
pietrolesci/small_unigramlm128k
Updated
pietrolesci/small_tokmix128k
Updated
pietrolesci/small_bpe128k
Updated
pietrolesci/small_multigram128k
Updated
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
datasets
56
pietrolesci/unimixlm
Viewer
•
Updated
•
81.9M
•
98
pietrolesci/me-minipile-evals
Viewer
•
Updated
•
1.22M
•
14
pietrolesci/pile-deduped
Viewer
•
Updated
•
748M
•
347
pietrolesci/pythia-deduped-memorisation-profiles
Viewer
•
Updated
•
2.13M
•
51
pietrolesci/pile-validation
Viewer
•
Updated
•
429k
•
151
pietrolesci/pile-deduped-subset
Viewer
•
Updated
•
16.3k
•
10
pietrolesci/pythia-deduped-stats
Viewer
•
Updated
•
16.3M
•
11
pietrolesci/pythia-deduped-stats-raw
Viewer
•
Updated
•
14.9M
•
1.93k
pietrolesci/agnews
Viewer
•
Updated
•
510k
•
57
pietrolesci/amazoncat-13k
Viewer
•
Updated
•
5.99M
•
136
•
1