[dependabot]

Bumps the ml-stack group with 2 updates in the / directory: sentence-transformers and transformers.

Updates sentence-transformers from 5.7.0 to 6.1.0

Release notes

Sourced from sentence-transformers's releases.

v6.1.0 - Documentation and input handling improvements

This is a small release with refreshed inference benchmarks, clearer documentation, and a few improvements to multimodal input handling.

Install this version with

# Training + Inference
pip install sentence-transformers[train]==6.1.0
Inference only, use one of:
pip install sentence-transformers==6.1.0
pip install sentence-transformers[onnx-gpu]==6.1.0
pip install sentence-transformers[onnx]==6.1.0
pip install sentence-transformers[openvino]==6.1.0
Multimodal dependencies (optional):
pip install sentence-transformers[image]==6.1.0
pip install sentence-transformers[audio]==6.1.0
pip install sentence-transformers[video]==6.1.0
Or combine as needed:
pip install sentence-transformers[train,onnx,image]==6.1.0

Changes

• Refreshed the inference benchmarks and efficiency guide, clarified benchmark labels, and documented ONNX CPU thread tuning (#3977, #4007, #3987).
• Added a shared input formats guide and SPARSEUP to the pretrained sparse model tables (#4036, #4038).
• Improved multimodal input handling and chat-template argument routing (#4037, #4048).
• Corrected loss formulas and argument descriptions in docstrings, clarified custom-module import requirements, and fixed links, typos, and optional type annotations.

For existing multimodal dict inputs, message content now follows each input's key order, which can change embeddings or scores. Video-frame URLs without recognizable image extensions need an explicit frame wrapper to remain one video. See #4037 for details.

All Changes

• [docs] Document that optional imports in a custom module become load-time requirements by @​abtonmoy in huggingface/sentence-transformers#3922
• [docs] Name the metric on the quality axis of the backend benchmark figures by @​tomaarsen in huggingface/sentence-transformers#3977
• docs: fix Constrastive/heaqp typos and subject-verb agreement by @​simpleqt in huggingface/sentence-transformers#3983
• docs: drop stale margin/squared Args from triplet loss docstrings by @​simpleqt in huggingface/sentence-transformers#3981
• docs: fix cross-encoder link and 'show case' wording in examples READMEs by @​simpleqt in huggingface/sentence-transformers#3982
• Correct the inverted CoSENT loss formula in two docstrings by @​vineethsaivs in huggingface/sentence-transformers#3984
• Fix implicit Optional annotations by @​VenishPaneliya in huggingface/sentence-transformers#3990
• [docs] Clarify sparse COO inputs for semantic_search by @​tomaarsen in huggingface/sentence-transformers#4001
• [docs] Refresh inference benchmarks and promote efficiency guides by @​tomaarsen in huggingface/sentence-transformers#4007
• [docs] Describe the MarginMSE target as the signed margin by @​MaxFreedomPollard in huggingface/sentence-transformers#4009
• Document ONNX CPU thread tuning and benchmark settings by @​ErnisBadmae in huggingface/sentence-transformers#3987
• docs: Document shared input formats across model archetypes by @​tomaarsen in huggingface/sentence-transformers#4036
• [docs] Add SPARSEUP to pretrained sparse models by @​tomaarsen in huggingface/sentence-transformers#4038
• Support multiple items per modality in one input by @​tomaarsen in huggingface/sentence-transformers#4037
• Fix load_audio_from_video routing in chat templates by @​tomaarsen in huggingface/sentence-transformers#4048

... (truncated)

Commits

• b0bf399 Release v6.1.0
• 55f2f9c Fix load_audio_from_video routing in chat templates (#4048)
• ad8e77b Support multiple items per modality in one input (#4037)
• ad356eb [docs] Add SPARSEUP to pretrained sparse models (#4038)
• 5f1bfb0 docs: Document shared input formats across model archetypes (#4036)
• b0ef5ef Document ONNX CPU thread tuning and benchmark settings (#3987)
• 6a841ce [docs] Describe the MarginMSE target as the signed margin (#4009)
• dc0d7c3 [docs] Refresh inference benchmarks and promote efficiency guides (#4007)
• 1a4b62a [docs] Clarify sparse COO inputs for semantic_search (#4001)
• 24d718f Fix implicit Optional annotations (#3990)
• Additional commits viewable in compare view

Updates transformers from 5.18.0 to 5.20.1

Release notes

Sourced from transformers's releases.

Release v5.20.1

Special surprise including the new model Beam from Reflection AI 🥳

Beam is a token-efficient, frontier-scale model focused on coding and agentic use cases. It is built for teams that want to run a powerful model on their own hardware, infrastructure, or cloud, and customize it to their specific needs.

Links: Documentation

• Fix small bug in auto mappings (#49664) by @​yonigozlan in #49664

Some real patch fixes

We broke loading models on older torch versions 😢 should be fixed now. But here is the full list of fixes

• Fix HunYuan-VL image assignment during batched generation (#49557) by @​Arnav66692
• Fix DeepGEMM Mega MoE under expert parallelism(#49428) by @​IlyasMoutawwakil
• Fix from_pretrained on torch 2.7: require torch>=2.8 for distributed checkpointing (#49652) by @​naveecode
• Make DtensorShardOperation work on torch 2.6 and 2.7 (#49657) by @​hazelxue

Full Changelog: huggingface/transformers@v5.20.0...v5.20.1

Release v5.20.0

New Model additions

Muse Spark Open

Muse Spark Open is a multimodal mixture-of-experts model from Meta that takes text, images, audio, and video as input and generates text. It has 816B total parameters (42B active per token) and a context window of up to 1,048,576 tokens shared across modalities. It is trained for agentic coding tasks such as multi-file editing, debugging, and tool use. Checkpoints are available in BF16 and NVFP4 formats.

Links: Documentation | Blog Post

• Add Muse Spark Open (#49644) by @​guarin in #49644

Ling 3.0

Ling 3.0 Flash is a hybrid Mixture-of-Experts language model from InclusionAI. Its decoder groups five Kimi Delta Attention (KDA) layers with one Multi-head Latent Attention (MLA) layer. The MLA blocks use interleaved rotary embeddings and a head-wise output gate, and the feed-forward blocks combine routed and shared experts. It is natively supported in Transformers as BailingHybridForCausalLM, so it loads without trust_remote_code=True.

Links: Documentation

• [Model] Add native Ling 3.0 Flash support (#49446) by @​icavan in #49446

Kolibri1

Kolibri 1 is Aleph Alpha's mixture-of-experts reasoning model, focused on German and English. Each of its 50 layers routes every token to 6 of 384 experts using sigmoid routing, and adds one shared expert. Four sliding-window attention layers alternate with one full-attention layer, and RoPE is applied only in the sliding-window layers.

... (truncated)

Commits

• 0ea8779 v5.20.1
• de4f35f Fix small bug in auto mappings (#49664)
• 759c538 Make DtensorShardOperation work on torch 2.6 and 2.7 (#49657)
• c180e2d Fix from_pretrained on torch 2.7: require torch>=2.8 for distributed checkpoi...
• 10d36aa Fix DeepGEMM Mega MoE under expert parallelism (#49428)
• e45b214 Fix HunYuan-VL image assignment during batched generation (#49557)
• 48e4bcd v5.20.0
• 487dcfa Add Muse Spark Open (#49644)
• d2e102b [docs] Hardware docs template (#49575)
• 9d59c2c Add MOSS-Transcribe-Diarize model (#48386)
• Additional commits viewable in compare view