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
Official PyTorch implementation of "AASIST: Audio Anti-Spoofing using Integrated Spectro-Temporal Graph Attention Networks"

AASIST This repository provides the overall framework for training and evaluating audio anti-spoofing systems proposed in 'AASIST: Audio Anti-Spoofing

Clova AI Research 56 Jan 02, 2023
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
deep-prae

Deep Probabilistic Accelerated Evaluation (Deep-PrAE) Our work presents an efficient rare event simulation methodology for black box autonomy using Im

Safe AI Lab 4 Apr 17, 2021
Implementation of ML models like Decision tree, Naive Bayes, Logistic Regression and many other

ML_Model_implementaion Implementation of ML models like Decision tree, Naive Bayes, Logistic Regression and many other dectree_model: Implementation o

Anshuman Dalai 3 Jan 24, 2022
A hybrid framework (neural mass model + ML) for SC-to-FC prediction

The current workflow simulates brain functional connectivity (FC) from structural connectivity (SC) with a neural mass model. Gradient descent is applied to optimize the parameters in the neural mass

Yilin Liu 1 Jan 26, 2022
Code for MentorNet: Learning Data-Driven Curriculum for Very Deep Neural Networks

MentorNet: Learning Data-Driven Curriculum for Very Deep Neural Networks This is the code for the paper: MentorNet: Learning Data-Driven Curriculum fo

Google 302 Dec 23, 2022
A chemical analysis of lipophilicities & molecule drawings including ML

A chemical analysis of lipophilicity & molecule drawings including a bit of ML analysis. This is a simple project that includes two Jupyter files (one

Aurimas A. Nausėdas 7 Nov 22, 2022
[ICCV 2021 Oral] Mining Latent Classes for Few-shot Segmentation

Mining Latent Classes for Few-shot Segmentation Lihe Yang, Wei Zhuo, Lei Qi, Yinghuan Shi, Yang Gao. This codebase contains baseline of our paper Mini

Lihe Yang 66 Nov 29, 2022
Dynamic Graph Event Detection

DyGED Dynamic Graph Event Detection Get Started pip install -r requirements.txt TODO Paper link to arxiv, and how to cite. Twitter Weather dataset tra

Mert Koşan 3 May 09, 2022
BboxToolkit is a tiny library of special bounding boxes.

BboxToolkit is a light codebase collecting some practical functions for the special-shape detection, such as oriented detection

jbwang1997 73 Jan 01, 2023
Official implementation of "Learning Forward Dynamics Model and Informed Trajectory Sampler for Safe Quadruped Navigation" (RSS 2022)

Intro Official implementation of "Learning Forward Dynamics Model and Informed Trajectory Sampler for Safe Quadruped Navigation" Robotics:Science and

Yunho Kim 21 Dec 07, 2022
Lunar is a neural network aimbot that uses real-time object detection accelerated with CUDA on Nvidia GPUs.

Lunar Lunar is a neural network aimbot that uses real-time object detection accelerated with CUDA on Nvidia GPUs. About Lunar can be modified to work

Zeyad Mansour 276 Jan 07, 2023
Callable PyTrees and filtered JIT/grad transformations => neural networks in JAX.

Equinox Callable PyTrees and filtered JIT/grad transformations = neural networks in JAX Equinox brings more power to your model building in JAX. Repr

Patrick Kidger 909 Dec 30, 2022
This repository contains the code for TACL2021 paper: SummaC: Re-Visiting NLI-based Models for Inconsistency Detection in Summarization

SummaC: Summary Consistency Detection This repository contains the code for TACL2021 paper: SummaC: Re-Visiting NLI-based Models for Inconsistency Det

Philippe Laban 24 Jan 03, 2023
Official Python implementation of the FuzionCoin protocol

PyFuzc Official Python implementation of the FuzionCoin protocol WARNING: Under construction. Use at your own risk. Some functions may not work. Setup

FuzionCoin 3 Jul 07, 2022
Composable transformations of Python+NumPy programsComposable transformations of Python+NumPy programs

Chex Chex is a library of utilities for helping to write reliable JAX code. This includes utils to help: Instrument your code (e.g. assertions) Debug

DeepMind 506 Jan 08, 2023
Guiding evolutionary strategies by (inaccurate) differentiable robot simulators @ NeurIPS, 4th Robot Learning Workshop

Guiding Evolutionary Strategies by Differentiable Robot Simulators In recent years, Evolutionary Strategies were actively explored in robotic tasks fo

Vladislav Kurenkov 4 Dec 14, 2021
Automatic tool focused on deriving metallicities of open clusters

metalcode Automatic tool focused on deriving metallicities of open clusters. Based on the method described in Pöhnl & Paunzen (2010, https://ui.adsabs

2 Dec 13, 2021
This is the research repository for Vid2Doppler: Synthesizing Doppler Radar Data from Videos for Training Privacy-Preserving Activity Recognition.

Vid2Doppler: Synthesizing Doppler Radar Data from Videos for Training Privacy-Preserving Activity Recognition This is the research repository for Vid2

Future Interfaces Group (CMU) 26 Dec 24, 2022
Spline is a tool that is capable of running locally as well as part of well known pipelines like Jenkins (Jenkinsfile), Travis CI (.travis.yml) or similar ones.

Welcome to spline - the pipeline tool Important note: Since change in my job I didn't had the chance to continue on this project. My main new project

Thomas Lehmann 29 Aug 22, 2022