Dariush Wahdany

Pattern guides/regex patterns don't seem to have quite the impact one would expect ```python from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("Neko-Institute-of-Science/LLaMA-7B-HF") model = AutoModelForCausalLM.from_pretrained("Neko-Institute-of-Science/LLaMA-7B-HF") llama = guidance.llms.Transformers(model=model, tokenizer=tokenizer, device=5)...

Fix BatchMemoryManager length

3

## Types of changes - [x] Bug fix (non-breaking change which fixes an issue) - [ ] New feature (non-breaking change which adds functionality) - [ ] Breaking change (fix...

`BatchSplittingSampler` return wrong length

2

## 🐛 Bug `BatchSplittingSampler` reports the length as ```python expected_batch_size = self.sampler.sample_rate * self.sampler.num_samples return int(len(self.sampler) * (expected_batch_size / self.max_batch_size)) ``` Converting the result simply to `int` leads to the...

feat: update lightning example to lightning 2.0

7

## Types of changes - [ ] Bug fix (non-breaking change which fixes an issue) - [ ] New feature (non-breaking change which adds functionality) - [ ] Breaking change...

Can't set parameters from API

1

I might have overlooked something, but it seems there is no way to adjust the parameters from API calls, e.g. you can't call `get_title_from_file(path, algo='max2')`.

feat: pass args via run

1

Fixes #33 This should be the most minimal-invasive way of passing arguments to pdftitle when calling it from another python module. It allows the module to be used in conjunction...

fix: mlflow version check

1

fix #167 by utilising `packaging.version`

The BaseIterator checks the mlflow version with `mlflow.__version__ < "2.2.1"` (see code below). This check does not function as intented, since for the latest version (2.10.0) the check is true,...

### Bug description If you pass a `torch.utils.data.dataloader.DataLoader` object as part of your hyperparameters and save it using in your `LightningModule` using `self.save_hyperparameters()`, it will hang `trainer.fit` ### What version...

Implementation of the paper **Beyond the Mean: Differentially Private Prototypes for Private Transfer Learning** from Dariush Wahdany, Matthew Jagielski, Adam Dziedzic, Franziska Boenisch. https://arxiv.org/abs/2406.08039