1e3ddcb2d0
* bug fixes
* organize imports
* wrap cpu warning in reference_compile
* Avoid needing repad_logits_with_grad, always repad with grads when training
I'm not 100% that the conditional with "or labels is None" makes sense though - not sure what the intention is there. Perhaps we can remove that?
* Revert "Avoid needing repad_logits_with_grad, always repad with grads when training"
This reverts commit cedcb4e89b.
* Fix grammar: keep -> keeps
* Propagate grammar fix with modular_model_converter
---------
Co-authored-by: Tom Aarsen <Cubiegamedev@gmail.com>
Co-authored-by: Tom Aarsen <37621491+tomaarsen@users.noreply.github.com>
5.3 KiB
ModernBERT
Overview
The ModernBERT model was proposed in Smarter, Better, Faster, Longer: A Modern Bidirectional Encoder for Fast, Memory Efficient, and Long Context Finetuning and Inference by Benjamin Warner, Antoine Chaffin, Benjamin Clavié, Orion Weller, Oskar Hallström, Said Taghadouini, Alexis Galalgher, Raja Bisas, Faisal Ladhak, Tom Aarsen, Nathan Cooper, Grifin Adams, Jeremy Howard and Iacopo Poli.
It is a refresh of the traditional encoder architecture, as used in previous models such as BERT and RoBERTa.
It builds on BERT and implements many modern architectural improvements which have been developed since its original release, such as:
- Rotary Positional Embeddings to support sequences of up to 8192 tokens.
- Unpadding to ensure no compute is wasted on padding tokens, speeding up processing time for batches with mixed-length sequences.
- GeGLU Replacing the original MLP layers with GeGLU layers, shown to improve performance.
- Alternating Attention where most attention layers employ a sliding window of 128 tokens, with Global Attention only used every 3 layers.
- Flash Attention to speed up processing.
- A model designed following recent The Case for Co-Designing Model Architectures with Hardware, ensuring maximum efficiency across inference GPUs.
- Modern training data scales (2 trillion tokens) and mixtures (including code ande math data)
The abstract from the paper is the following:
Encoder-only transformer models such as BERT offer a great performance-size tradeoff for retrieval and classification tasks with respect to larger decoder-only models. Despite being the workhorse of numerous production pipelines, there have been limited Pareto improvements to BERT since its release. In this paper, we introduce ModernBERT, bringing modern model optimizations to encoder-only models and representing a major Pareto improvement over older encoders. Trained on 2 trillion tokens with a native 8192 sequence length, ModernBERT models exhibit state-of-the-art results on a large pool of evaluations encompassing diverse classification tasks and both single and multi-vector retrieval on different domains (including code). In addition to strong downstream performance, ModernBERT is also the most speed and memory efficient encoder and is designed for inference on common GPUs.
The original code can be found here.
Resources
A list of official Hugging Face and community (indicated by 🌎) resources to help you get started with ModernBert.
- A notebook on how to finetune for General Language Understanding Evaluation (GLUE) with Transformers, also available as a Google Colab notebook. 🌎
- A script on how to finetune for text similarity or information retrieval with Sentence Transformers. 🌎
- A script on how to finetune for information retrieval with PyLate. 🌎
ModernBertConfig
autodoc ModernBertConfig
ModernBertModel
autodoc ModernBertModel - forward
ModernBertForMaskedLM
autodoc ModernBertForMaskedLM - forward
ModernBertForSequenceClassification
autodoc ModernBertForSequenceClassification - forward
ModernBertForTokenClassification
autodoc ModernBertForTokenClassification - forward