githubnemo
daf335f503
The current state of the PEFT docs is not one of structure and I was constantly annoyed that whenever I wanted to change something there were several places that needed touching and they all felt disconnected. So this is my attempt at structuring the docs. Some of these ideas are quite old (discussed in 01/2025) but are still valid. I've removed most of the code guides without replacement. That's not ideal, I think we should have code examples but I'm think they should be method-focused. Maybe one general example of a training workflow is sufficient because most methods follow the same scheme. All details from the method guides (prompting, lora, oft/boft, etc.) are now integrated into the respective method pages instead. I would have hesitated to do this if these guides would have integrated information about the adapters but they didn't. I think it makes a lot more sense to have one place for each method to gather examples/tips/recommendations and that is now the `package_refernce/<method>` page. This page now also hosts a small space that shows the MetaMathQA (and potentially other) benchmark results highlighted for that method. I've moved the LoRA initializations to `package_reference/lora#Initialization` and converted the init methods to `<hfoption>`-tags. This collapses them to a list but may reduce searchability through the document - at least firefox is not able to search 'through' the option tabs. This also doesn't make them appear in the ToC and people specifically searching for, say, PiSSA won't find it directly. I think that's OK though, since the search is able to locate it. The quicktour is a bit more detailed about what happens under the hood (quick doesn't have to mean simplistic) and includes some new visualizations. I hope that we can integrate more visualizations in the future where it makes sense. * Remove PEFT method space + front page buttons The space was not that useful anymore since most methods are compatible with most models. The front page buttons are, at least temporarily, with the exception of the quicktour and method overview buttons. I like the visuals but there should only be elements that are useful. --------- Co-authored-by: Benjamin Bossan <BenjaminBossan@users.noreply.github.com> Co-authored-by: Steven Liu <59462357+stevhliu@users.noreply.github.com>
3.2 KiB
DeLoRA: Decoupled Low-rank Adaptation
DeLoRA is a parameter-efficient fine-tuning technique that implicitly maintains a Frobenius boundary with respect to the pretrained weights by normalizing and scaling learnable low-rank matrices. This effectively decouples the learning of directions (BA term) and magnitude (boundary term) of the weight updates, avoiding catastrophic shifts in the adapted weights and enhancing robustness to hyperparameter choices.
Note:
- use a learning rate 10-100x larger than for standard LoRA variants (typical values from 1e-3/1e-2/..)
- ensure the initial boundary parameter lambda is not too small (typical values around 10/15/..). Setting different lambdas to different layers is possible
DeLoRA currently has the following constraints:
- Only nn.Linear layers are supported.
- Quantized layers are not supported.
If these constraints don't work for your use case, consider other methods instead.
The abstract from the paper is:
Parameter-Efficient FineTuning (PEFT) methods have recently gained significant popularity thanks to the widespread availability of large-scale pretrained models. These methods allow for quick adaptation to downstream tasks with minimal computational cost. However, popular finetuning methods such as LoRA exhibit limited robustness when it comes to hyperparameter choices or extended training regimes, preventing optimal out-of-the-box performance. In contrast, bounded approaches, such as ETHER, provide greater robustness but are limited to extremely low-rank adaptations and fixed-strength transformations, reducing their adaptation expressive power. In this work, we propose Decoupled Low-rank Adaptation (DeLoRA), a novel finetuning method that normalizes and scales learnable low-rank matrices. By bounding the distance of the transformation, DeLoRA effectively decouples the angular learning from the adaptation strength, enhancing robustness without compromising performance. Through evaluations on subject-driven image generation, natural language understanding, and instruction tuning, we show that DeLoRA matches or surpasses performance of competing PEFT methods, while exhibiting stronger robustness.
Benchmark overview
API
DeloraConfig
autodoc tuners.delora.config.DeloraConfig
DeloraModel
autodoc tuners.delora.model.DeloraModel