Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Upgrade seqio to upstream to workaround the MaxText dependency issue …
…for inference (#1225) Default version built with MaxText image is `seqio==0.0.16` which introduces error ``` 0: Traceback (most recent call last): 0: File "/opt/maxtext/MaxText/inference_microbenchmark.py", line 26, in <module> 0: from jetstream.engine import token_utils 0: File "/opt/jetstream/jetstream/engine/token_utils.py", line 24, in <module> 0: from seqio.vocabularies import SentencePieceVocabulary 0: File "/usr/local/lib/python3.12/dist-packages/seqio/__init__.py", line 18, in <module> 0: from seqio.dataset_providers import * 0: File "/usr/local/lib/python3.12/dist-packages/seqio/dataset_providers.py", line 60, in <module> 0: @dataclasses.dataclass(frozen=True) 0: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 0: File "/usr/lib/python3.12/dataclasses.py", line 1258, in wrap 0: return _process_class(cls, init, repr, eq, order, unsafe_hash, 0: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 0: File "/usr/lib/python3.12/dataclasses.py", line 994, in _process_class 0: cls_fields.append(_get_field(cls, name, type, kw_only)) 0: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 0: File "/usr/lib/python3.12/dataclasses.py", line 852, in _get_field 0: raise ValueError(f'mutable default {type(f.default)} for field ' 0: ValueError: mutable default <class 'seqio.vocabularies.PassThroughVocabulary'> for field vocabulary is not allowed: use default_factory ``` Upgrade seqio to directly use upstream repo to workaround the issue. (This is not the perfect but workable solution) --------- Co-authored-by: Vladislav Kozlov <vkozlov@nvidia.com>
- Loading branch information