ai-agent

pytorch_influence_functions: Reveal Why Models Predict

Name: pytorch_influence_functions: Reveal Why Models Predict
Author: nimarb

A Python library for founders building AI products who need to trace predictions back to training data, using the ICML‑2017 influence‑function method.

From the repo · “This is a PyTorch reimplementation of Influence Functions from the ICML2017 best paper: Understanding Black-box Predictions via Influence Functions by Pang Wei Koh and Percy Liang.”

348 stars74 forksPythonGuide quality 9/10 · reviewed Oct 10Updated 10/29/2023100% free · open source

What it does

Pytorch_influence_functions helps startup founders understand and analyze the predictions made by their deep learning models by identifying the most influential training data points

When to use it

•When you need to debug and understand why your model is making certain predictions
•When you want to identify biases in your training data and their impact on model predictions
•When you need to explain model decisions to stakeholders or customers

Ready-to-paste prompt

python examples/influence_functions_example.py --dataset cifar10 --model resnet18 --num_samples 100

Heads up: Make sure you have PyTorch installed and the version is compatible with the pytorch_influence_functions library, as it requires PyTorch 1.9 or later

Saves to your device

Use with Claude

New

Skip the builder — one click puts this in Claude, Cursor, Antigravity and more.

✅ Light setup

Installs with a command or two; your AI agent can do it for you.

Try it instantly — no install

Claude Code

mkdir -p ~/.claude/skills/pytorch-influence-functions && curl -fsSL https://workflowstacks.com/api/skills/pytorch-influence-functions/claude-skill -o ~/.claude/skills/pytorch-influence-functions/SKILL.md

Open in another AI app

Opens the app with this repo with the prompt ready to go — no copy-paste needed.

Connect the whole catalog (MCP)

claude mcp add --transport http workflowstacks https://workflowstacks.com/api/mcp

Adds a WorkflowStacks connector to Claude Code: search and load any skill here by chatting.

How pytorch_influence_functions: Reveal Why Models Predict works

Codeflow

Free to inspect

Pytorch_influence_functions: Reveal Why Models Predict is a small Python project (~1k lines across 7 code files). Setup is light: installs like a normal app. Reading the code is optional. Last commit 36 months ago, Other license, has a test suite.

Size

Small codebase

~1k lines · 7 code files · 8 min skim

Setup

One-command install

Installs like a normal app. Reading the code is optional.

Runs on

Python

No API keys detected

Python 100%

Where to start reading

1
README.md
Start here — what it does and how to install it
2
pytorch_influence_functions/__init__.py
Inside pytorch_influence_functions/ — the main logic begins here
3
examples/test_influence_functions.py
A worked example — copy this to get going

What's in each folder

pytorch_influence_functions/Core code — the actual logic4 files

examples/Examples you can copy2 files

figs/Folder3 files

READMEHas testsThin docsNo CIExamples includedOther licenseLast update 36 mo ago