dependabot[bot] opened a new pull request, #39312: URL: https://github.com/apache/beam/pull/39312
Bumps [transformers](https://github.com/huggingface/transformers) from 4.55.4 to 5.5.0. <details> <summary>Release notes</summary> <p><em>Sourced from <a href="https://github.com/huggingface/transformers/releases";>transformers's releases</a>.</em></p> <blockquote> <h1>Release v5.5.0</h1> <!-- raw HTML omitted --> <h2>New Model additions</h2> <h3>Gemma4</h3> <p><a href="https://github.com/huggingface/transformers/blob/HEAD/INSET_PAPER_LINK";>Gemma 4</a> 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.</p> <!-- raw HTML omitted --> <p>You can find all the original Gemma 4 checkpoints under the <a href="https://huggingface.co/collections/google/gemma-4-release-67c6c6f89c4f76621268bb6d";>Gemma 4</a> release.</p> <p>The key difference from previous Gemma releases is the new design to process <strong>images of different sizes</strong> using a <strong>fixed-budget number of tokens</strong>. 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:</p> <ul> <li>The total number of pixels must fit within a patch budget</li> <li>Both height and width must be divisible by <strong>48</strong> (= patch size 16 × pooling kernel 3)</li> </ul> <blockquote> <p>[!IMPORTANT] Gemma 4 does <strong>not</strong> apply the standard ImageNet mean/std normalization that many other vision models use. The model's own patch embedding layer handles the final scaling internally (shifting values to the [-1, 1] range).</p> </blockquote> <p>The number of "soft tokens" (aka vision tokens) an image processor can produce is configurable. The supported options are outlined below and the default is <strong>280 soft tokens</strong> per image.</p> <table> <thead> <tr> <th align="center">Soft Tokens</th> <th align="center">Patches (before pooling)</th> <th align="center">Approx. Image Area</th> </tr> </thead> <tbody> <tr> <td align="center">70</td> <td align="center">630</td> <td align="center">~161K pixels</td> </tr> <tr> <td align="center">140</td> <td align="center">1,260</td> <td align="center">~323K pixels</td> </tr> <tr> <td align="center"><strong>280</strong></td> <td align="center"><strong>2,520</strong></td> <td align="center"><strong>~645K pixels</strong></td> </tr> <tr> <td align="center">560</td> <td align="center">5,040</td> <td align="center">~1.3M pixels</td> </tr> <tr> <td align="center">1,120</td> <td align="center">10,080</td> <td align="center">~2.6M pixels</td> </tr> </tbody> </table> <p>To encode positional information for each patch in the image, Gemma 4 uses a learned 2D position embedding table. The position table stores up to 10,240 positions per axis, which allows the model to handle very large images. Each position is a learned vector of the same dimensions as the patch embedding. The 2D RoPE which Gemma 4 uses independently rotate half the attention head dimensions for the x-axis and the other half for the y-axis. This allows the model to understand spatial relationships like "above," "below," "left of," and "right of."</p> <h3>NomicBERT</h3> <p>NomicBERT is a BERT-inspired encoder model that applies Rotary Position Embeddings (RoPE) to create reproducible long context text embeddings. It is the first fully reproducible, open-source text embedding model with 8192 context length that outperforms both OpenAI Ada-002 and OpenAI text-embedding-3-small on short-context MTEB and long context LoCo benchmarks. The model generates dense vector embeddings for various tasks including search, clustering, and classification using specific instruction prefixes.</p> <p><strong>Links:</strong> <a href="https://huggingface.co/docs/transformers/main/en/model_doc/nomic_bert";>Documentation</a> | <a href="https://arxiv.org/abs/2402.01613";>Paper</a></p> <ul> <li>Internalise the NomicBERT model (<a href="https://redirect.github.com/huggingface/transformers/issues/43067";>#43067</a>) by <a href="https://github.com/ed22699";><code>@ed22699</code></a> in <a href="https://redirect.github.com/huggingface/transformers/pull/43067";>#43067</a></li> </ul> <h3>MusicFlamingo</h3> <p>Music Flamingo is a fully open large audio–language model designed for robust understanding and reasoning over music. It builds upon the Audio Flamingo 3 architecture by including Rotary Time Embeddings (RoTE), which injects temporal position information to enable the model to handle audio sequences up to 20 minutes. The model features a unified audio encoder across speech, sound, and music with special sound boundary tokens for improved audio sequence modeling.</p> <p><strong>Links:</strong> <a href="https://huggingface.co/docs/transformers/main/en/model_doc/musicflamingo";>Documentation</a> | <a href="https://huggingface.co/papers/2511.10289";>Paper</a></p> <ul> <li>Add Music Flamingo (<a href="https://redirect.github.com/huggingface/transformers/issues/43538";>#43538</a>) by <a href="https://github.com/lashahub";><code>@lashahub</code></a> in <a href="https://redirect.github.com/huggingface/transformers/pull/43538";>#43538</a></li> </ul> <!-- raw HTML omitted --> </blockquote> <p>... (truncated)</p> </details> <details> <summary>Commits</summary> <ul> <li><a href="https://github.com/huggingface/transformers/commit/c1c34249fa27deefbd4a377dfbf883a39baf5c6d";><code>c1c3424</code></a> update</li> <li><a href="https://github.com/huggingface/transformers/commit/20bff6865a756a074f5b893b57f0ae438b25ec46";><code>20bff68</code></a> update release workflow</li> <li><a href="https://github.com/huggingface/transformers/commit/89564412a56ae6581f8aa48a533a835860dc9f43";><code>8956441</code></a> v5.5.0</li> <li><a href="https://github.com/huggingface/transformers/commit/5135e5efa7203cd23aac0866de12dfeef038422d";><code>5135e5e</code></a> casually dropping the most capable open weights on the planet (<a href="https://redirect.github.com/huggingface/transformers/issues/45192";>#45192</a>)</li> <li><a href="https://github.com/huggingface/transformers/commit/a594e09e3924120f1f5508e7d81946bf3504df2b";><code>a594e09</code></a> Internalise the NomicBERT model (<a href="https://redirect.github.com/huggingface/transformers/issues/43067";>#43067</a>)</li> <li><a href="https://github.com/huggingface/transformers/commit/4932e9721e230bea915341e7f04db32885b6c6af";><code>4932e97</code></a> Fix resized LM head weights being overwritten by post_init (<a href="https://redirect.github.com/huggingface/transformers/issues/45079";>#45079</a>)</li> <li><a href="https://github.com/huggingface/transformers/commit/57e84139542c8c297873f35fcd25f66ffcf132ae";><code>57e8413</code></a> [Qwen3.5 MoE] Add _tp_plan to ForConditionalGeneration (<a href="https://redirect.github.com/huggingface/transformers/issues/45124";>#45124</a>)</li> <li><a href="https://github.com/huggingface/transformers/commit/b10552e99dc4974b30126995baea455df43f8476";><code>b10552e</code></a> Fix TypeError: 'NoneType' object is not iterable in GenerationMixin.generate ...</li> <li><a href="https://github.com/huggingface/transformers/commit/423f2a31d2bd05bdc1dc30dd938389edaa998fde";><code>423f2a3</code></a> fix(models): Fix dtype mismatch in SwitchTransformers and TimmWrapperModel (#...</li> <li><a href="https://github.com/huggingface/transformers/commit/ade7a05a42bf53b183bb78c181743be063c5ff14";><code>ade7a05</code></a> Generalize gemma vision mask to videos (<a href="https://redirect.github.com/huggingface/transformers/issues/45185";>#45185</a>)</li> <li>Additional commits viewable in <a href="https://github.com/huggingface/transformers/compare/v4.55.4...v5.5.0";>compare view</a></li> </ul> </details> <br /> [](https://docs.github.com/en/github/managing-security-vulnerabilities/about-dependabot-security-updates#about-compatibility-scores) 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-automerge-start) [//]: # (dependabot-automerge-end) --- <details> <summary>Dependabot commands and options</summary> <br /> 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 this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself) - `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself) - `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself) You can disable automated security fix PRs for this repo from the [Security Alerts page](https://github.com/apache/beam/network/alerts). </details> -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
