⬆️ dep-bump(deps): Bump the dependency-packages group across 1 directory with 3 updates

[dependabot]

Bumps the dependency-packages group with 3 updates in the / directory: datasets, transformers and lm-eval.

Updates datasets from 3.2.0 to 4.8.4

Release notes

Sourced from datasets's releases.

4.8.4

What's Changed

• Support latest torchvision by @​lhoestq in huggingface/datasets#8087
• fix regression when loading JSON with one file = one object by @​lhoestq in huggingface/datasets#8086

Full Changelog: huggingface/datasets@4.8.3...4.8.4

4.8.3

What's Changed

• Fix split_dataset_by_node step by @​lhoestq in huggingface/datasets#8081
• Fix docstring of Json.cast_storage by @​albertvillanova in huggingface/datasets#8080

Full Changelog: huggingface/datasets@4.8.2...4.8.3

4.8.2

What's Changed

• Json type for empty struct by @​lhoestq in huggingface/datasets#8074

Full Changelog: huggingface/datasets@4.8.1...4.8.2

4.8.1

What's Changed

• Fix formatted iter arrow double yield by @​HaukurPall in huggingface/datasets#8063

Full Changelog: huggingface/datasets@4.8.0...4.8.1

4.8.0

Dataset Features

• Read (and write) from HF Storage Buckets: load raw data, process and save to Dataset Repos by @​lhoestq in huggingface/datasets#8064

  from datasets import load_dataset
  # load raw data from a Storage Bucket on HF
  ds = load_dataset("buckets/username/data-bucket", data_files=["*.jsonl"])
  # or manually, using hf:// paths
  ds = load_dataset("json", data_files=["hf://buckets/username/data-bucket/*.jsonl"])
  # process, filter
  ds = ds.map(...).filter(...)
  # publish the AI-ready dataset
  ds.push_to_hub("username/my-dataset-ready-for-training")

  This also fixes multiprocessed push_to_hub on macos that was causing segfault (now it uses spawn instead of fork). And it bumps dill and multiprocess versions to support python 3.14

• Datasets streaming iterable packaged improvements and fixes by @​Michael-RDev in huggingface/datasets#8068

  • added max_shard_size to IterableDataset.push_to_hub (but requires iterating twice to know the full dataset twice - improvements are welcome)

... (truncated)

Commits

• a0ce369 release: 4.8.4 (#8088)
• 0856825 fix regression when loading JSON with one file = one object (#8086)
• 6ca75d0 Support latest torchvision (#8087)
• 4a0c9c9 set dev version (#8083)
• d4942e2 Release 4.8.3 (#8082)
• 1b31309 Fix split_dataset_by_node step (#8081)
• 4a4813f Fix docstring of Json.cast_storage (#8080)
• 68294ab set dev version (#8077)
• 2c46ab1 release 4.8.2 (#8076)
• 97ba494 Json type for empty struct (#8074)
• Additional commits viewable in compare view

Updates transformers from 4.46.0 to 5.5.4

Release notes

Sourced from transformers's releases.

Patch release v5.5.4

This is mostly some fixes that are good to have asap, mostly for tokenizers; ** Fix Kimi-K2.5 tokenizer regression and _patch_mistral_regex Attribute… (#45305) by ArthurZucker

For training: ** Fix #45305 + add regression test GAS (#45349) by florian6973, SunMarc ** Fix IndexError with DeepSpeed ZeRO-3 when kernels rotary is active (#…) by ArthurZucker

And for Qwen2.5-VL : ** Fix Qwen2.5-VL temporal RoPE scaling applied to still images (#45330) by Kash6, zucchini-nlp

Patch release: v5.5.3

Small patch release to fix device_map support for Gemma4! It contains the following commit:

• [gemma4] Fix device map auto (#45347) by @​Cyrilvallez

Patch release: v5.5.2

Small patch dedicated to optimizing gemma4, fixing inference with use_cache=False due to k/v states sharing between layers, as well as conversion mappings for some models that would inconsistently serialize their weight names. It contains the following PRs:

• Add MoE to Gemma4 TP plan (#45219) by @​sywangyi and @​Cyrilvallez
• [gemma4] Dissociate kv states sharing from the Cache (#45312) by @​Cyrilvallez
• [gemma4] Remove all shared weights, and silently skip them during loading (#45336) by @​Cyrilvallez
• Fix conversion mappings for vlms (#45340) by @​Cyrilvallez

Patch release v5.5.1

This patch is very small and focuses on vLLM and Gemma4!

** Fix export for gemma4 and add Integration tests (#45285) by @​Cyrilvallez ** Fix vllm cis (#45139) by @​ArthurZucker

Release v5.5.0

New Model additions

Gemma4

Gemma 4 is a multimodal model with pretrained and instruction-tuned variants, available in 1B, 13B, and 27B parameters. The architecture is mostly the same as the previous Gemma versions. The key differences are a vision processor that can output images of fixed token budget and a spatial 2D RoPE to encode vision-specific information across height and width axis.

You can find all the original Gemma 4 checkpoints under the Gemma 4 release.

The key difference from previous Gemma releases is the new design to process images of different sizes using a fixed-budget number of tokens. Unlike many models that squash every image into a fixed square (like 224×224), Gemma 4 keeps the image's natural aspect ratio while making it the right size. There a a couple constraints to follow:

• The total number of pixels must fit within a patch budget
• Both height and width must be divisible by 48 (= patch size 16 × pooling kernel 3)

... (truncated)

Commits

• 75d3bdc v5.5.4
• 285dcec Fix IndexError with DeepSpeed ZeRO-3 when kernels rotary is active (#45414)
• 16f5c6c Fix Qwen2.5-VL temporal RoPE scaling applied to still images (#45330)
• 1b5a1d9 Fix #45305 + add regression test GAS (#45349)
• fdffefb Fix Kimi-K2.5 tokenizer regression and _patch_mistral_regex AttributeError (#...
• c6c8503 Release: v5.5.3
• dccd162 [gemma4] Fix device map auto (#45347)
• 52d24b0 Release: 5.5.2
• 5df3452 [gemma4] Remove all shared weights, and silently skip them during loading (#4...
• 2cf8ff0 Fix conversion mappings for vlms (#45340)
• Additional commits viewable in compare view

Updates lm-eval from 0.4.5 to 0.4.11

Release notes

Sourced from lm-eval's releases.

v0.4.11 Release Notes

Minor release. Stay tuned for bigger changes next release.

New Platform Support

• Windows ML Backend — Native Windows ML inference support by @​chapsiru and @​chemwolf6922 in #3470, #3564, #3565

New Benchmarks & Tasks

• BEAR knowledge probe by @​plonerma in #3496

Task Version Changes

The following tasks have updated versions. Results from a previous task versions may not be directly comparable. See the linked PRs or individual task READMEs for changelogs.

afrobench_belebele (all variants): 2 → 3 in #3551 evalita_llm: 0.0 → 0.1 in #3551 include (all 90 language variants): 0.0 → 0.1 in #3551 mgsm_direct (all 11 language variants): 3.0 → 4.0 by @​LakshyaChaudhry in #3574

Fixes & Improvements

• Fixed SQuAD v2 evaluation by @​HydrogenSulfate in #3535
• Fixed MasakhaNEWS tasks — replaced non-existent headline_text field with headline by @​Mr-Neutr0n in #3567
• Fixed incorrect task configs by @​baberabb in #3552
• Replaced eval() with ast.literal_eval in task configs for safer parsing by @​baberabb in #3577
• Fixed SGLang duplicate registration error by @​enpimashin in #3543
• Restored hf_transfer import check by @​baberabb in #3563
• Fixed modify_gen_kwargs call in vLLM VLMs by @​hmellor in #3573
• Refactored vLLM gen_kwargs normalization inline to modify_gen_kwargs; fixed cached gen_kwargs mutation by @​baberabb in #3582
• Fixed README for task-listing CLI command by @​UltimateJupiter in #3545
• Updated dependencies by @​baberabb in #3546

New Contributors

• @​HydrogenSulfate made their first contribution in EleutherAI/lm-evaluation-harness#3535
• @​UltimateJupiter made their first contribution in EleutherAI/lm-evaluation-harness#3545
• @​enpimashin made their first contribution in EleutherAI/lm-evaluation-harness#3543
• @​chapsiru made their first contribution in EleutherAI/lm-evaluation-harness#3470
• @​chemwolf6922 made their first contribution in EleutherAI/lm-evaluation-harness#3565
• @​plonerma made their first contribution in EleutherAI/lm-evaluation-harness#3496
• @​hmellor made their first contribution in EleutherAI/lm-evaluation-harness#3573
• @​Mr-Neutr0n made their first contribution in EleutherAI/lm-evaluation-harness#3567
• @​LakshyaChaudhry made their first contribution in EleutherAI/lm-evaluation-harness#3574

Full Changelog: EleutherAI/lm-evaluation-harness@v0.4.10...v0.4.11

v0.4.10

Highlights

... (truncated)

Commits

• 27988a2 increment version (#3585)
• 413588c refactor(vllm): inline gen_kwargs normalization to modify_gen_kwargs; fix...
• 7418f1c fix: use answer_number directly in mgsm_direct doc_to_target (#3574)
• eccfc5b ignore ruff rule S101 (assert-used) (#3579)
• 10462f9 fix: replace non-existent 'headline_text' with 'headline' in MasakhaNEWS task...
• a5b407d pacify pre-commit (#3577)
• e45984c fix(winml): fix winml incompatible parameters (#3575)
• b177b36 Fix call to modify_gen_kwargs in vllm_vlms.py (#3573)
• 69145a0 Adding the BEAR knowledge probe (#3496)
• 8c17e10 removed unused device parameter for winml model (#3564)
• Additional commits viewable in compare view

Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase.

---

Dependabot commands and options

You can trigger Dependabot actions by commenting on this PR:

• @dependabot rebase will rebase this PR
• @dependabot recreate will recreate this PR, overwriting any edits that have been made to it
• @dependabot show <dependency name> ignore conditions will show all of the ignore conditions of the specified dependency
• @dependabot ignore <dependency name> major version will close this group update PR and stop Dependabot creating any more for the specific dependency's major version (unless you unignore this specific dependency's major version or upgrade to it yourself)
• @dependabot ignore <dependency name> minor version will close this group update PR and stop Dependabot creating any more for the specific dependency's minor version (unless you unignore this specific dependency's minor version or upgrade to it yourself)
• @dependabot ignore <dependency name> will close this group update PR and stop Dependabot creating any more for the specific dependency (unless you unignore this specific dependency or upgrade to it yourself)
• @dependabot unignore <dependency name> will remove all of the ignore conditions of the specified dependency
• @dependabot unignore <dependency name> <ignore condition> will remove the ignore condition of the specified dependency and ignore conditions