dmis-lab/UltraBERT-380M
fill-mask model by dmis-lab
Older snapshot: check the original source for current information.
From the publisher
Model card & documentation
Source previewRead the publisher’s intended use, setup instructions, evaluations and limitations. The original model card is the source of truth.
UltraBERT is a family of English bidirectional encoders with a native 32K context that reach state of the art on both classification and retrieval, two task families on which previous encoders had to choose a side. For more information about UltraBERT, please check out our paper. UltraBERT is supported natively in transformers >= 5.20. On earlier 5.x versions (tested with 5.16.1), pass trustremotecode=True. FlashAttention 2 is recommended for long inputs; SDPA and eager attention are also supported. The checkpoint loads into the usual task classes: We recommend starting from the settings used in the paper. For classification: AdamW (β1 0.9, β2 0.98, ε 1e-6), weight decay 0.1, a linear schedule with 5% warm-up, and a learning rate swept over {5e-6,8e-6,1e-5,3e-5,5e-5,8e-5}. For dense retrieval: mean pooling with a contrastive loss and a wider learning-rate sweep of {1e-5,3e-5,5e-5,8e-5,1e-4,2e-4,3e-4,5e-4,8e-4,1e-3}. Every model is fine-tuned per task with the same grid search. For the retrieval…
Read the full model card ↗ · Preview checked 2026-10-30T12:27:29.500Z
Inside the original model card — Document outline
- UltraBERT-380M
- ✨ Highlights
- 🚀 Usage
- Fine-Tuning
- 📊 Evaluation
- Base Size (at most 200M parameters)
- Medium Size (at most 500M parameters)
- Long-Context Passkey Retrieval
- 🏛️ Model Description
- What Is in This Checkpoint
- ⚙️ Pre-Training
- Limitations
Headings are captured from the source. Links open the publisher’s document, not a locally hosted copy.
Links from the model card
References supplied by the publisher, not independently verified endorsements. Check the destination before downloading files or entering credentials.
Documentation belongs to its respective authors. Reported project/model license: apache-2.0. A listing is not a grant of reuse or training rights. Confirm the document’s own terms at the source.
Model overview
dmis-lab/UltraBERT-380M is a fill-mask model repository published by dmis-lab on Hugging Face. The source reports the transformers library.
This page summarizes Hub metadata. For intended use, training data, evaluation results and limitations, consult the original model card.
Model facts
- Task
- fill-mask
- Library
- transformers
- Recent downloads (30 days)
- 8,162
- Cumulative likes
- 66
- Hugging Face trending score
- 56
- Reported safetensors parameters
- 429,966,336
- Architecture
- UltrabertForMaskedLM
- License
- apache-2.0
- Access
- Not gated by Hugging Face
- Created
- 2026-10-10T10:53:18.000Z
- Last modified
- 2026-10-16T18:11:46.000Z
Compatibility and lineage
Reported languages: en
Base-model lineage was not included.
Parameter count is not a RAM/VRAM requirement. Check precision, quantization, context length and runtime compatibility in the model card; no hardware or API-cost claim is inferred here.
Use and evaluate
Check license terms and access requirements first. Review the model card’s documented loading instructions, evaluations and safety limitations. Benchmark scores are not imported by this directory, and popularity is not an accuracy ranking.
No model weights are downloaded or executed by AltAPIs. Never enable remote model code without reviewing it.
Source and freshness
Source: Hugging Face Hub. Metadata observed 2026-10-21T06:43:12.081Z. Daily imports are snapshots, not real-time monitoring.
Popularity and source listings do not establish security, suitability, licensing rights or benchmark performance.