* fix(dataprep): smart_chunk_text single-chunk path leaks internal tensor type when eos_token_id is None
RawTextDataLoader.smart_chunk_text()'s single-chunk branch only
converts `tokens` to a plain Python list inside the
`if eos_token_id is not None:` guard. When a tokenizer has no
eos_token_id configured, that conversion is skipped entirely and the
function returns whatever internal tensor-like object came out of
the tokenizer normalization step (e.g. a torch.Tensor) as
"input_ids", instead of a list of ints.
The sibling multi-chunk branch a few lines below does the conversion
unconditionally, before checking eos_token_id -- the two branches of
the same method disagree on output type depending purely on whether
the tokenizer has an EOS token. Downstream, create_causal_dataset()
does `labels = [list(ids) for ids in input_ids]`; list()'ing a
tensor produces a list of 0-d tensor elements rather than plain
ints, inconsistent with every multi-chunk sample and liable to break
type inference in Dataset.from_dict()/downstream collation.
Fix: move the list conversion out of the eos_token_id guard,
matching the multi-chunk branch's existing pattern.
Added test_smart_chunk_text_single_chunk_no_eos_returns_plain_list
to tests/test_raw_text.py, confirmed red against unfixed code
(assertion failure: input_ids was a MockTensor, not a list) and
green after the fix. Full tests/test_raw_text.py (both test
functions) passes. ruff check + the repo's ruff-format-with-kwargs
script: clean.
Note: tests/test_raw_text.py does not appear to be wired into any
.github/workflows/*.yml CI job (a pre-existing repo characteristic,
not something introduced by this change) -- verified locally via
`python3 tests/test_raw_text.py`.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Etherl <61019402+Etherll@users.noreply.github.com>
c2762f7f42