Modernize + verify 5 tutorial notebooks (Longformer/IMDB, node2vec, bikeshare, jigsaw) - #2

Open

jlealtru wants to merge 18 commits into

mainfrom

modernize-notebooks

Commits on May 17, 2026

Modernize repo: uv env + device-portable RoBERTa+IMDB
jlealtru
committed
e083bdb

Commits on Jul 13, 2026

Modernize + verify 5 tutorial notebooks (Longformer/IMDB, node2vec, bikeshare, jigsaw)

jlealtru
and
claude
committed
3881c85
Add data-fetch helper scripts; restore README references

jlealtru
and
claude
committed
545f1f3
Smoke-verify BigBird + RoBERTa on MPS; add SMOKE_TEST toggle

jlealtru
and
claude
committed
685b3fd
Verify + modernize the two ETM (Pitchfork) notebooks

jlealtru
and
claude
committed
b8d9251
Address review: _utils import path, BigBird block-sparse, Jigsaw smoke sampling

jlealtru
and
claude
committed
ce84cac

Commits on Jul 14, 2026

Fix ETM cache collision between the two Pitchfork notebooks

jlealtru
and
claude
committed
8cccb2a
Make MPS-era dataloader/precision settings device-conditional

jlealtru
and
claude
committed
7b3a88d
Add optional spacy-cuda group for GPU tokenization (Linux/NVIDIA)

jlealtru
and
claude
committed
81733a3

Commits on Jul 15, 2026

Speed up ETM training with a GPU-resident dense BOW matrix

jlealtru
and
claude
committed
0a00092

Commits on Jul 17, 2026

Tokenizer-only spaCy pipeline for the ETM spacy notebook

jlealtru
and
claude
committed
7724f67
Terser notebook comments; add topic exploration to the spacy ETM

jlealtru
and
claude
committed
df51bee
Follow the paper's bow normalization for the encoder input

jlealtru
and
claude
committed
c8792d4

Commits on Jul 18, 2026

Commits on Aug 5, 2026

Add an EmbeddingGemma arm to the BERTopic Pitchfork notebook

jlealtru
and
claude
committed
10fb757
Use the current eval_strategy name in the RoBERTa notebook prose

jlealtru
and
claude
committed
a319156
Re-run the BERTopic Pitchfork notebook and encode both models up front

jlealtru
and
claude
committed
d80fa04