A library for finding knowledge neurons in pretrained transformer models.

Overview

knowledge-neurons

An open source repository replicating the 2021 paper Knowledge Neurons in Pretrained Transformers by Dai et al., and extending the technique to autoregressive models, as well as MLMs.

The Huggingface Transformers library is used as the backend, so any model you want to probe must be implemented there.

Currently integrated models:

BERT_MODELS = ["bert-base-uncased", "bert-base-multilingual-uncased"]
GPT2_MODELS = ["gpt2"]
GPT_NEO_MODELS = [
    "EleutherAI/gpt-neo-125M",
    "EleutherAI/gpt-neo-1.3B",
    "EleutherAI/gpt-neo-2.7B",
]

The technique from Dai et al. has been used to locate knowledge neurons in the huggingface bert-base-uncased model for all the head/relation/tail entities in the PARAREL dataset. Both the neurons, and more detailed results of the experiment are published at bert_base_uncased_neurons/*.json and can be replicated by running pararel_evaluate.py. More details in the Evaluations on the PARAREL dataset section.

Setup

Either clone the github, and run scripts from there:

git clone knowledge-neurons
cd knowledge-neurons

Or install as a pip package:

pip install knowledge-neurons

Usage & Examples

An example using bert-base-uncased:

from knowledge_neurons import KnowledgeNeurons, initialize_model_and_tokenizer, model_type
import random

# first initialize some hyperparameters
MODEL_NAME = "bert-base-uncased"

# to find the knowledge neurons, we need the same 'facts' expressed in multiple different ways, and a ground truth
TEXTS = [
    "Sarah was visiting [MASK], the capital of france",
    "The capital of france is [MASK]",
    "[MASK] is the capital of france",
    "France's capital [MASK] is a hotspot for romantic vacations",
    "The eiffel tower is situated in [MASK]",
    "[MASK] is the most populous city in france",
    "[MASK], france's capital, is one of the most popular tourist destinations in the world",
]
TEXT = TEXTS[0]
GROUND_TRUTH = "paris"

# these are some hyperparameters for the integrated gradients step
BATCH_SIZE = 20
STEPS = 20 # number of steps in the integrated grad calculation
ADAPTIVE_THRESHOLD = 0.3 # in the paper, they find the threshold value `t` by multiplying the max attribution score by some float - this is that float.
P = 0.5 # the threshold for the sharing percentage

# setup model & tokenizer
model, tokenizer = initialize_model_and_tokenizer(MODEL_NAME)

# initialize the knowledge neuron wrapper with your model, tokenizer and a string expressing the type of your model ('gpt2' / 'gpt_neo' / 'bert')
kn = KnowledgeNeurons(model, tokenizer, model_type=model_type(MODEL_NAME))

# use the integrated gradients technique to find some refined neurons for your set of prompts
refined_neurons = kn.get_refined_neurons(
    TEXTS,
    GROUND_TRUTH,
    p=P,
    batch_size=BATCH_SIZE,
    steps=STEPS,
    coarse_adaptive_threshold=ADAPTIVE_THRESHOLD,
)

# suppress the activations at the refined neurons + test the effect on a relevant prompt
# 'results_dict' is a dictionary containing the probability of the ground truth being generated before + after modification, as well as other info
# 'unpatch_fn' is a function you can use to undo the activation suppression in the model. 
# By default, the suppression is removed at the end of any function that applies a patch, but you can set 'undo_modification=False', 
# run your own experiments with the activations / weights still modified, then run 'unpatch_fn' to undo the modifications
results_dict, unpatch_fn = kn.suppress_knowledge(
    TEXT, GROUND_TRUTH, refined_neurons
)

# suppress the activations at the refined neurons + test the effect on an unrelated prompt
results_dict, unpatch_fn = kn.suppress_knowledge(
    "[MASK] is the official language of the solomon islands",
    "english",
    refined_neurons,
)

# enhance the activations at the refined neurons + test the effect on a relevant prompt
results_dict, unpatch_fn = kn.enhance_knowledge(TEXT, GROUND_TRUTH, refined_neurons)

# erase the weights of the output ff layer at the refined neurons (replacing them with zeros) + test the effect
results_dict, unpatch_fn = kn.erase_knowledge(
    TEXT, refined_neurons, target=GROUND_TRUTH, erase_value="zero"
)

# erase the weights of the output ff layer at the refined neurons (replacing them with an unk token) + test the effect
results_dict, unpatch_fn = kn.erase_knowledge(
    TEXT, refined_neurons, target=GROUND_TRUTH, erase_value="unk"
)

# edit the weights of the output ff layer at the refined neurons (replacing them with the word embedding of 'target') + test the effect
# we can make the model think the capital of france is London!
results_dict, unpatch_fn = kn.edit_knowledge(
    TEXT, target="london", neurons=refined_neurons
)

for bert models, the position where the "[MASK]" token is located is used to evaluate the knowledge neurons, (and the ground truth should be what the mask token is expected to be), but due to the nature of GPT models, the last position in the prompt is used by default, and the ground truth is expected to immediately follow.

In GPT models, due to the subword tokenization, the integrated gradients are taken n times, where n is the length of the expected ground truth in tokens, and the mean of the integrated gradients at each step is taken.

for bert models, the ground truth is currently expected to be a single token. Multi-token ground truths are on the todo list.

Evaluations on the PARAREL dataset

To ensure that the repo works correctly, figures 3 and 4 from the knowledge neurons paper are reproduced below. In general the results appear similar, except suppressing unrelated facts appears to have a little more of an affect in this repo than in the paper's original results.*

Below are Dai et al's, and our result, respectively, for suppressing the activations of the refined knowledge neurons in pararel: knowledge neuron suppression / dai et al. knowledge neuron suppression / ours

And Dai et al's, and our result, respectively, for enhancing the activations of the knowledge neurons: knowledge neuron enhancement / dai et al. knowledge neuron enhancement / ours

To find the knowledge neurons in bert-base-uncased for the PARAREL dataset, and replicate figures 3. and 4. from the paper, you can run

# find knowledge neurons + test suppression / enhancement (this will take a day or so on a decent gpu) 
# you can skip this step since the results are provided in `bert_base_uncased_neurons`
python -m torch.distributed.launch --nproc_per_node=NUM_GPUS_YOU_HAVE pararel_evaluate.py
# plot results 
python plot_pararel_results.py

*It's unclear where the difference comes from, but my suspicion is they made sure to only select facts with different relations, whereas in the plots below, only a different pararel UUID was selected. In retrospect, this could actually express the same fact, so I'll rerun these experiments soon.

TODO:

  • Better documentation
  • Publish PARAREL results for bert-base-multilingual-uncased
  • Publish PARAREL results for bert-large-uncased
  • Publish PARAREL results for bert-large-multilingual-uncased
  • Multiple masked tokens for bert models
  • Find good dataset for GPT-like models to evaluate knowledge neurons (PARAREL isn't applicable since the tail entities aren't always at the end of the sentence)
  • Add negative examples for getting refined neurons (i.e expressing a different fact in the same way)
  • Look into different attribution methods (cf. https://arxiv.org/pdf/2010.02695.pdf)

Citations

@article{Dai2021KnowledgeNI,
  title={Knowledge Neurons in Pretrained Transformers},
  author={Damai Dai and Li Dong and Y. Hao and Zhifang Sui and Furu Wei},
  journal={ArXiv},
  year={2021},
  volume={abs/2104.08696}
}
Owner
EleutherAI
EleutherAI
Adaptive Dropblock Enhanced GenerativeAdversarial Networks for Hyperspectral Image Classification

This repo holds the codes of our paper: Adaptive Dropblock Enhanced GenerativeAdversarial Networks for Hyperspectral Image Classification, which is ac

Feng Gao 17 Dec 28, 2022
[CVPR22] Official codebase of Semantic Segmentation by Early Region Proxy.

RegionProxy Figure 2. Performance vs. GFLOPs on ADE20K val split. Semantic Segmentation by Early Region Proxy Yifan Zhang, Bo Pang, Cewu Lu CVPR 2022

Yifan 54 Nov 29, 2022
Pytorch implementation of Zero-DCE++

Zero-DCE++ You can find more details here: https://li-chongyi.github.io/Proj_Zero-DCE++.html. You can find the details of our CVPR version: https://li

Chongyi Li 157 Dec 23, 2022
Open source repository for the code accompanying the paper 'PatchNets: Patch-Based Generalizable Deep Implicit 3D Shape Representations'.

PatchNets This is the official repository for the project "PatchNets: Patch-Based Generalizable Deep Implicit 3D Shape Representations". For details,

16 May 22, 2022
Pytorch implementation of CoCon: A Self-Supervised Approach for Controlled Text Generation

COCON_ICLR2021 This is our Pytorch implementation of COCON. CoCon: A Self-Supervised Approach for Controlled Text Generation (ICLR 2021) Alvin Chan, Y

alvinchangw 79 Dec 18, 2022
Rewrite ultralytics/yolov5 v6.0 opencv inference code based on numpy, no need to rely on pytorch

Rewrite ultralytics/yolov5 v6.0 opencv inference code based on numpy, no need to rely on pytorch; pre-processing and post-processing using numpy instead of pytroch.

炼丹去了 21 Dec 12, 2022
Implementation of ConvMixer-Patches Are All You Need? in TensorFlow and Keras

Patches Are All You Need? - ConvMixer ConvMixer, an extremely simple model that is similar in spirit to the ViT and the even-more-basic MLP-Mixer in t

Sayan Nath 8 Oct 03, 2022
This is an official implementation for "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows" on Semantic Segmentation.

Swin Transformer for Semantic Segmentation of satellite images This repo contains the supported code and configuration files to reproduce semantic seg

23 Oct 10, 2022
Robotics environments

Robotics environments Details and documentation on these robotics environments are available in OpenAI's blog post and the accompanying technical repo

Farama Foundation 121 Dec 28, 2022
Code for the paper titled "Prabhupadavani: A Code-mixed Speech Translation Data for 25 languages"

Prabhupadavani: A Code-mixed Speech Translation Data for 25 languages Code for the paper titled "Prabhupadavani: A Code-mixed Speech Translation Data

Ayush Daksh 12 Dec 01, 2022
Pytorch modules for paralel models with same architecture. Ideal for multi agent-based systems

WideLinears Pytorch parallel Neural Networks A package of pytorch modules for fast paralellization of separate deep neural networks. Ideal for agent-b

1 Dec 17, 2021
We utilize deep reinforcement learning to obtain favorable trajectories for visual-inertial system calibration.

Unified Data Collection for Visual-Inertial Calibration via Deep Reinforcement Learning Update: The lastest code will be updated in this branch. Pleas

ETHZ ASL 27 Dec 29, 2022
unofficial pytorch implementation of RefineGAN

RefineGAN unofficial pytorch implementation of RefineGAN (https://arxiv.org/abs/1709.00753) for CSMRI reconstruction, the official code using tensorpa

xinby17 5 Jul 21, 2022
Repository for "Toward Practical Monocular Indoor Depth Estimation" (CVPR 2022)

Toward Practical Monocular Indoor Depth Estimation Cho-Ying Wu, Jialiang Wang, Michael Hall, Ulrich Neumann, Shuochen Su [arXiv] [project site] DistDe

Meta Research 122 Dec 13, 2022
TeST: Temporal-Stable Thresholding for Semi-supervised Learning

TeST: Temporal-Stable Thresholding for Semi-supervised Learning TeST Illustration Semi-supervised learning (SSL) offers an effective method for large-

Xiong Weiyu 1 Jul 14, 2022
Code for ACL 2019 Paper: "COMET: Commonsense Transformers for Automatic Knowledge Graph Construction"

To run a generation experiment (either conceptnet or atomic), follow these instructions: First Steps First clone, the repo: git clone https://github.c

Antoine Bosselut 575 Jan 01, 2023
Neural network-based build time estimation for additive manufacturing

Neural network-based build time estimation for additive manufacturing Oh, Y., Sharp, M., Sprock, T., & Kwon, S. (2021). Neural network-based build tim

Yosep 1 Nov 15, 2021
the official code for ICRA 2021 Paper: "Multimodal Scale Consistency and Awareness for Monocular Self-Supervised Depth Estimation"

G2S This is the official code for ICRA 2021 Paper: Multimodal Scale Consistency and Awareness for Monocular Self-Supervised Depth Estimation by Hemang

NeurAI 4 Jul 27, 2022
A flexible ML framework built to simplify medical image reconstruction and analysis experimentation.

meddlr Getting Started Meddlr is a config-driven ML framework built to simplify medical image reconstruction and analysis problems. Installation To av

Arjun Desai 36 Dec 16, 2022
Implementation of Google Brain's WaveGrad high-fidelity vocoder

WaveGrad Implementation (PyTorch) of Google Brain's high-fidelity WaveGrad vocoder (paper). First implementation on GitHub with high-quality generatio

Ivan Vovk 363 Dec 27, 2022