Make differentially private training of transformers easy for everyone

Overview

private-transformers

This codebase facilitates fast experimentation of differentially private training of Hugging Face transformers.


What is this? Why an extra codebase?

  • This codebase provides a privacy engine that builds off Opacus, but works way more smoothly with Hugging Face's transformers library.
  • Additionally, we support the ghost clipping technique (see Section 4 of this preprint on how it works) which allows privately training large transformers with considerably reduced memory cost -- in many cases, almost as light as non-private training -- at a modest run-time overhead.
  • With this codebase, we have fine-tuned very large pretrained models, yielding some of the best performing differentially private NLP models to date. Some of these models have performance matching strong non-private baseline approaches. We see strong empirical evidence that highly performant DP NLP models could be built on modest datasets.

Installation

Make sure you have python>=3.8; run the following command:

pip install git+https://github.com/lxuechen/private-transformers.git

To check the package is installed properly, be sure to run the test suite (requires pytest and a GPU) via the following command:

pytest -s tests

Usage

Basic usage

Privately training Hugging Face transformers with our codebase simply consists of 4 steps:

  1. Create your favourite transformer model and optimizer; attach this optimizer to a PrivacyEngine
  2. Compute a per-example loss (1-D tensor) for a mini-batch of data
  3. Pass the loss to optimizer.step or optimizer.virtual_step as a keyword argument
  4. Repeat from step 2

Below is a quick example:

import transformers, torch
from private_transformers import PrivacyEngine
import torch.nn.functional as F

device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
model = transformers.GPT2LMHeadModel.from_pretrained('distilgpt2').to(device)
optimizer = torch.optim.Adam(params=model.parameters(), lr=1e-4)
privacy_engine = PrivacyEngine(
    model,
    batch_size=10,
    sample_size=50000,
    epochs=3,
    max_grad_norm=0.1,
    target_epsilon=3,
)
privacy_engine.attach(optimizer)

batch_size, seq_len = 10, 20
# Inputs are batch-first format, i.e., the first dimension of tensors must be batch dimension.
input_ids = torch.randint(size=[batch_size, seq_len], low=0, high=100, device=device)
# Calling `.train()` is very important; otherwise underlying forward and backward hooks don't run.
model.train()
outputs = model(input_ids=input_ids, return_dict=True)
labels = input_ids[:, 1:, ]
logits = outputs.logits[:, :-1, :].permute(0, 2, 1)
# `loss` is a 1-D tensor of shape (batch_size,).
loss = F.cross_entropy(logits, labels, reduction="none").mean(dim=1)
# This step is different from existing workflows: 
#   Don't call `loss.backward`; leave it to `optimizer.step` to handle backward.
optimizer.step(loss=loss)

The biggest differences compared to Opacus are:

  • We require the per-example loss (a 1-D tensor) be passed into optimizer.step (or optimizer.virtual_step)
  • The per-example loss must be passed in as a keyword argument.
  • loss.backward() shouldn't be called on the user end; it's called internally in optimizer.step ( or optimizer.virtual_step).
  • Inputs should be in batch-first format; there isn't a toggle to switch between different formats in the engine.

Ghost clipping: memory saving differentially private learning

Turning on ghost clipping requires changing only 1 line. You should notice a drastic reduction in peak GPU memory usage once this is turned on, at a potential cost of slower training speed. One might find this especially useful when constrained to only use older GPUs with small VRAMs or fitting super large models.

import transformers, torch
from private_transformers import PrivacyEngine

device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
model = transformers.GPT2LMHeadModel.from_pretrained('distilgpt2').to(device)
optimizer = torch.optim.Adam(params=model.parameters(), lr=1e-4)
privacy_engine = PrivacyEngine(
    model,
    batch_size=10,
    sample_size=50000,
    epochs=3,
    max_grad_norm=0.1,
    target_epsilon=3,
    ghost_clipping=True,  # The only change you need to make!
)
privacy_engine.attach(optimizer)

We ran stringent numerical tests to ensure the double-backward implementation is correct. Check out files in the tests folder for more on this.

Examples

Code in the examples folder roughly reproduces our results for the table-to-text and classification tasks. There may be some minor discrepancies, since hyperparameters there aren't exactly what's used in the paper. Nevertheless, it should be sufficient to get things started. Detailed instructions are in the readme file of each subfolder.

Currently supported Hugging Face models

Not all models in the Hugging Face library are supported. The main additional work here is to

  1. support per-example gradients for bespoke modules (e.g., T5LayerNorm), and
  2. ensure position_ids are repeated.

We plan to support more models in the future if there's such a need. Feel free to open an issue if you may want to try out specific models that aren't in the current list.

FAQ

I wrote some answers to potential questions here.

Acknowledgements

It would have been impossible to develop this codebase without cool past works and existing codebases. We roughly follow the PrivacyEngine design in Opacus==0.13.0. We directly use an off-the-shelf package for tightly tracking tradeoff functions while composing multiple private mechanisms.

Disclaimer

  • This codebase is not yet production-grade, e.g., cryptographically secure PRNGs are required for sampling noise -- our codebase currently does not use these strong PRNGs. This codebase also isn't immune to floating point representation attacks.
  • This codebase is born out of the need to experiment with various things for differentially private NLP in rapidly succession. I've tried my best to write clean code, though parts of this codebase may be less tidy than I had hoped given the extremely tight timeline.

Citation

If you found this codebase useful in your research, please consider citing:

@misc{li2021large,
      title={Large Language Models Can Be Strong Differentially Private Learners}, 
      author={Xuechen Li and Florian Tramรจr and Percy Liang and Tatsunori Hashimoto},
      year={2021},
      eprint={2110.05679},
      archivePrefix={arXiv},
      primaryClass={cs.LG}
}
Comments
  • Support BART model

    Support BART model

    Hi, I'm trying to apply your code in BART model. But I got the error like the below:

    ValueError: Ghost clipping does not support parameter sharing. Parameter sharing may be due to default parameter sharing between lm_head and embedding.Please use a model without parameter sharing for ghost clipping.
    

    Does it not support BART model yet??

    opened by SeolhwaLee 7
  • Set another seed won't change the result

    Set another seed won't change the result

    Hi Xuechen,

    I have another issue with the training seed. I would like to relax the random seed so that I can get some statistical results. Tried many different ways but even comment out the set_seed() function, the eva acc is the same until the last digit. May I ask how to relax the random seed? I'm doing experiments on examples/classification.

    Thanks!

    opened by JunyiZhu-AI 6
  • Customize loss function / adding regularizer under privacy setting?

    Customize loss function / adding regularizer under privacy setting?

    Hi, thanks again for the great work and the codebase!

    I have a question -- how I'd want to customize loss function in the codebase? I've been trying to do that, e.g. adding a per-example L1 regularization term to vector_loss in trainer, but I didn't manage to get it running after several attempts.

    There's a related discussion/PR in Opacus codebase https://github.com/pytorch/opacus/issues/249.

    However, there're a few tricky things I can see: -- In private-transformers, backward() behavior is not managed on the user end. -- also, 1-D vector_loss is required for private gradient update - optimizer.stepor optimizer.virtual_step

    My intuition is that I can add to vector_loss (per-example loss) at this line before the loss gets passed to the privacy engine.

    However I am afraid privacy concern is also an issue. I am aware of that Private-Transformers overrides compute_loss() in HF trainer, to exclude regularization terms that might mess up with privacy accounting.

    Sorry my question is not super detailed but I hope this makes sense and really appreciate for any comments.

    Thank you!

    opened by shi-kejian 4
  • Using dataloader with fixed batch size

    Using dataloader with fixed batch size

    Hi, thanks for providing this codebase!

    So for a while I've been using Opacus to experiment with DP-SGD and RoBERTa, but I wanted to check out your PrivacyEngine, mainly because of the training speed and memory optimizations. With Opacus, I always trained with their UniformWithReplacementSampler for accurate RDP accounting and as far as I can tell, you're training with fixed size batches in your examples. I'm wondering if there's a reason the UniformWithReplacementSampler isn't needed in your codebase anymore, and if the uniform sampler is compatible with your modified PrivacyEngine because the optimizer needs to be able to deal with variations in batch size?

    opened by xplip 4
  • How to set max_compositions

    How to set max_compositions

    Hi Chen, do you know how to set the max_compositions/steps param? The default at https://github.com/lxuechen/private-transformers/blob/684e27fcd9978539fbabc357c7ea506c0353c771/private_transformers/privacy_utils/privacy_engine.py#L148 is 0 but would raise an error

    private-transformers/private_transformers/privacy_utils/accounting/gdp_accounting.py:33: RuntimeWarning: invalid value encountered in double_scalars return norm.cdf(-eps / mu + mu / 2) - np.exp(eps) * norm.cdf(-eps / mu - mu / 2)
    rv_accountant/accountant.py:55: RuntimeWarning: divide by zero encountered in double_scalars mesh_size = 2*eps_error / np.sqrt(2*max_compositions*np.log(2/eta0))
    
    opened by hlzhang109 3
  • Setting small target epsilon like 0.1 fails training

    Setting small target epsilon like 0.1 fails training

    Hi, @lxuechen I tried to set epsilon as 0.1 on SST-2, but it results in a large noise_multiplier: 20853.95 and fails the training where the accuracy is near 0.5 However, setting epsilon as 1 works well. Any idea about this?

    opened by LinkToPast1900 3
  • Private gradient seemingly has been overwritten by non-private gradient.

    Private gradient seemingly has been overwritten by non-private gradient.

    Hi Xuechen, thanks for providing this codebase!

    I tried tweaking the code in examples/classification but the network does not perform as expected. In particular, I tried zeroing out all gradient by this command in _step() and _ghost_step() functions in privacy_engine.py:

    param.grad /= self.batch_size
    param.grad.mul_(0)
    

    After adding this multiplication the network has been trained as normally. And because with the same seed, the network has been even trained to give out the same eval acc. Could you reproduce this result at your private repo? If it behaves like this, then I suppose that the private gradient has been overwritten by the non-private one.

    opened by JunyiZhu-AI 3
  • Questions about sigma search and epsilon from composed tradeoff functions

    Questions about sigma search and epsilon from composed tradeoff functions

    (Making a new issue for this because you probably weren't notified of my comment in the closed original issue)

    Sorry for having to reopen this, but I do have two more (perhaps related) questions after all and would really appreciate if you could help clarify them.

    1. When using the automated sigma search (based on a specified target epsilon and N epochs), the final epsilon computed by the PrivacyEngine after training for N epochs is always much higher than the target epsilon, so it seems that the sigma chosen by get_sigma_from_rdp is too high. This also happens when I run the sentence classification and the table2test examples in the repo. E.g., instead of my target epsilon 8, I will end up with something like epsilon 10-11. How did you get your final epsilon to match the target epsilon in the experiments in your paper?

    2. How do you compute the converted epsilon from composed tradeoff functions when let's say training SST-2 with the default hyperparameters from the examples? Do you reduce the num_compositions=1000 in _eps_from_glw to something way lower than 1000 because the script only runs for ~400 optimization steps and would otherwise always throw the Numerical composition of tradeoff functions failed! Double check privacy parameters. error?

    Originally posted by @xplip in https://github.com/lxuechen/private-transformers/issues/7#issuecomment-987020758

    opened by xplip 3
  • What is the best way to handle large models?

    What is the best way to handle large models?

    Hi all, I was trying to fine-tune GPT-J 6B but I encounter Out Of Memory errors if I use a single-gpu, for non-private training I managed to solve them by using deepspeed but it seems that I cannot use that with opacus or with this codebase. Do you know how I could solve this problem? Thank you in advance:)

    opened by Pier297 2
  • No such file or directory

    No such file or directory

    I want to finetune qqp and here comes an error:

    File "/private-transformers-main/examples/classification/run_classification.py", line 545, in main train_dataset = FewShotDataset(data_args, tokenizer=tokenizer, mode="train", use_demo=use_demo) File "/private-transformers-main/examples/classification/src/dataset.py", line 377, in init with FileLock(lock_path): File "/home/anaconda3/envs/fuck/lib/python3.8/site-packages/filelock/_api.py", line 214, in enter self.acquire() File "/home/anaconda3/envs/fuck/lib/python3.8/site-packages/filelock/_api.py", line 170, in acquire self._acquire() File "/home/anaconda3/envs/fuck/lib/python3.8/site-packages/filelock/_unix.py", line 35, in _acquire fd = os.open(self._lock_file, open_mode) FileNotFoundError: [Errno 2] No such file or directory: 'classification/data/original/QQP/cached_train_RobertaTokenizer_256_qqp_few_shot.lock'

    how can I get this file? thanks.

    opened by trestad 2
  • [DistilBERT] RuntimeError: stack expects each tensor to be equal size

    [DistilBERT] RuntimeError: stack expects each tensor to be equal size

    Hi, @lxuechen, thanks for your repo.

    I met a problem as follows when I tied to finetune DistilBERT. Both BERT and Roberta work well. Any idea about this? Thanks!

    Traceback (most recent call last): ... File "/opt/conda/lib/python3.8/site-packages/torch/autograd/grad_mode.py", line 27, in decorate_context return func(*args, **kwargs) File "/opt/conda/lib/python3.8/site-packages/private_transformers/privacy_utils/privacy_engine.py", line 360, in step self._ghost_step(loss=kwargs.pop("loss")) File "/opt/conda/lib/python3.8/site-packages/private_transformers/privacy_utils/privacy_engine.py", line 261, in _ghost_step self._ghost_helper(loss) File "/opt/conda/lib/python3.8/site-packages/torch/autograd/grad_mode.py", line 27, in decorate_context return func(*args, **kwargs) File "/opt/conda/lib/python3.8/site-packages/private_transformers/privacy_utils/privacy_engine.py", line 334, in _ghost_helper coef_sample = self.get_coef_sample() File "/opt/conda/lib/python3.8/site-packages/private_transformers/privacy_utils/privacy_engine.py", line 348, in get_coef_sample norm_sample = self.get_norm_sample() File "/opt/conda/lib/python3.8/site-packages/private_transformers/privacy_utils/privacy_engine.py", line 343, in get_norm_sample norm_sample = torch.stack([param.norm_sample for name, param in self.named_params], dim=0).norm(2, dim=0) RuntimeError: stack expects each tensor to be equal size, but got [50] at entry 0 and [1] at entry 1

    (50 is my batch size)

    opened by LinkToPast1900 1
  • v0.3.0 fixes

    v0.3.0 fixes

    Non-structural fixes.

    • [ ] Convert to make_private style to avoid bad syntax highlighting during static analysis
    • [ ] Improve the cleanliness of examples
    • [ ] Refactor test file and use functorch to simplify ground truth gradients' logic
    • [ ] Don't compute per-sample gradients for weights which don't require gradients
    • [ ] Use the new smart resizer for tokenizer and model
    • [ ] Refactor decoding to use new left padding based construction
    opened by lxuechen 0
Releases(v0.2.3)
Owner
Xuechen Li
learning to learn
Xuechen Li
High performance, easy-to-use, and scalable machine learning (ML) package, including linear model (LR), factorization machines (FM), and field-aware factorization machines (FFM) for Python and CLI interface.

What is xLearn? xLearn is a high performance, easy-to-use, and scalable machine learning package that contains linear model (LR), factorization machin

Chao Ma 3k Jan 03, 2023
๐Ÿ“š Papermill is a tool for parameterizing, executing, and analyzing Jupyter Notebooks.

papermill is a tool for parameterizing, executing, and analyzing Jupyter Notebooks. Papermill lets you: parameterize notebooks execute notebooks This

nteract 5.1k Jan 03, 2023
FcaNet: Frequency Channel Attention Networks

FcaNet: Frequency Channel Attention Networks PyTorch implementation of the paper "FcaNet: Frequency Channel Attention Networks". Simplest usage Models

327 Dec 27, 2022
Minimal implementation and experiments of "No-Transaction Band Network: A Neural Network Architecture for Efficient Deep Hedging".

No-Transaction Band Network: A Neural Network Architecture for Efficient Deep Hedging Minimal implementation and experiments of "No-Transaction Band N

19 Jan 03, 2023
Code to train models from "Paraphrastic Representations at Scale".

Paraphrastic Representations at Scale Code to train models from "Paraphrastic Representations at Scale". The code is written in Python 3.7 and require

John Wieting 71 Dec 19, 2022
Code for the AAAI-2022 paper: Imagine by Reasoning: A Reasoning-Based Implicit Semantic Data Augmentation for Long-Tailed Classification

Imagine by Reasoning: A Reasoning-Based Implicit Semantic Data Augmentation for Long-Tailed Classification (AAAI 2022) Prerequisite PyTorch = 1.2.0 P

16 Dec 14, 2022
A self-supervised 3D representation learning framework named viewpoint bottleneck.

Pointly-supervised 3D Scene Parsing with Viewpoint Bottleneck Paper Created by Liyi Luo, Beiwen Tian, Hao Zhao and Guyue Zhou from Institute for AI In

63 Aug 11, 2022
Self-Supervised Learning with Data Augmentations Provably Isolates Content from Style

Self-Supervised Learning with Data Augmentations Provably Isolates Content from Style [NeurIPS 2021] Official code to reproduce the results and data p

Yash Sharma 27 Sep 19, 2022
CondenseNet: Light weighted CNN for mobile devices

CondenseNets This repository contains the code (in PyTorch) for "CondenseNet: An Efficient DenseNet using Learned Group Convolutions" paper by Gao Hua

Shichen Liu 690 Nov 30, 2022
(IEEE TIP 2021) Regularized Densely-connected Pyramid Network for Salient Instance Segmentation

RDPNet IEEE TIP 2021: Regularized Densely-connected Pyramid Network for Salient Instance Segmentation PyTorch training and testing code are available.

Yu-Huan Wu 41 Oct 21, 2022
Moer Grounded Image Captioning by Distilling Image-Text Matching Model

Moer Grounded Image Captioning by Distilling Image-Text Matching Model Requirements Python 3.7 Pytorch 1.2 Prepare data Please use git clone --recurse

YE Zhou 60 Dec 16, 2022
[์ œ 13ํšŒ ํˆฌ๋น…์Šค ์ปจํผ๋Ÿฐ์Šค] OK Mugle! - ์žฅ๋ฅด๋ถ€ํ„ฐ ๋ฉœ๋กœ๋””๊นŒ์ง€, Content-based Music Recommendation

Ok Mugle! ๐ŸŽต ์žฅ๋ฅด๋ถ€ํ„ฐ ๋ฉœ๋กœ๋””๊นŒ์ง€, Content-based Music Recommendation 'Ok Mugle!'์€ ์ œ13ํšŒ ํˆฌ๋น…์Šค ์ปจํผ๋Ÿฐ์Šค(2022.01.15)์—์„œ ์ง„ํ–‰ํ•œ ์Œ์•… ์ถ”์ฒœ ํ”„๋กœ์ ํŠธ์ž…๋‹ˆ๋‹ค. Description ๐Ÿ“– ๋ณธ ํ”„๋กœ์ ํŠธ์—์„œ๋Š” Kakao

SeongBeomLEE 5 Oct 09, 2022
A hue shift helper for OBS

obs-hue-shift A hue shift helper for OBS This is a repo based on the really nice script Hegemege made. The original script can be found https://gist.g

Alexis Tyler 1 Jan 10, 2022
This Repo is the official CUDA implementation of ICCV 2019 Oral paper for CARAFE: Content-Aware ReAssembly of FEatures

Introduction This Repo is the official CUDA implementation of ICCV 2019 Oral paper for CARAFE: Content-Aware ReAssembly of FEatures. @inproceedings{Wa

Jiaqi Wang 42 Jan 07, 2023
Official PyTorch Implementation of "Self-supervised Auxiliary Learning with Meta-paths for Heterogeneous Graphs". NeurIPS 2020.

Self-supervised Auxiliary Learning with Meta-paths for Heterogeneous Graphs This repository is the implementation of SELAR. Dasol Hwang* , Jinyoung Pa

MLV Lab (Machine Learning and Vision Lab at Korea University) 48 Nov 09, 2022
Toontown House CT Edition

Toontown House: Classic Toontown House Classic source that should just work. โ“ W

Open Source Toontown Servers 5 Jan 09, 2022
A Python Package for Portfolio Optimization using the Critical Line Algorithm

PyCLA A Python Package for Portfolio Optimization using the Critical Line Algorithm Getting started To use PyCLA, clone the repo and install the requi

19 Oct 11, 2022
ImageBART: Bidirectional Context with Multinomial Diffusion for Autoregressive Image Synthesis

ImageBART NeurIPS 2021 Patrick Esser*, Robin Rombach*, Andreas Blattmann*, Bjรถrn Ommer * equal contribution arXiv | BibTeX | Poster Requirements A sui

CompVis Heidelberg 110 Jan 01, 2023
[SIGGRAPH Asia 2021] Pose with Style: Detail-Preserving Pose-Guided Image Synthesis with Conditional StyleGAN

Pose with Style: Detail-Preserving Pose-Guided Image Synthesis with Conditional StyleGAN [Paper] [Project Website] [Output resutls] Official Pytorch i

Badour AlBahar 215 Dec 17, 2022
A Pytorch Implementation of Source Data-free Domain Adaptation for a Faster R-CNN

A Pytorch Implementation of Source Data-free Domain Adaptation for a Faster R-CNN Please follow Faster R-CNN and DAF to complete the environment confi

2 Jan 12, 2022