OSS Tanbou
View on GitHubProject website

Mammoth — benchmark and extend continual-learning methods on one PyTorch foundation

OSS scale score 163.1OSS health 56Update pace 0 commits/30d
About these scores

OSS scale score is an unbounded metric that log-compresses and weights Stars, Watchers, Forks, and Contributors. Discovery score is the current OSS scale score minus the score at discovery. Update pace is commits in the last 30 days, growth momentum is the OSS scale score difference within the recent observation window, and OSS health is a 0–100 rating based on available recency, Community Health, and release data.

Stars
838
Primary language
Python
License
MIT
Repository last updated
May 20, 2026

Overview

Mammoth provides a common PyTorch framework for implementing, benchmarking, and reproducing Continual Learning research. DER/DER++ and many other methods share one CLI and configuration model across numerous datasets, and new models or datasets can be added to the same structure.

Features and best fit

Based on official documentation; not hands-on tested · Content checked:

Compare many continual-learning methods through one CLI

More than 70 models can be launched through common main.py arguments and configuration while switching datasets and experimental settings.

Sources: [2][3]

Extend models, datasets, and training strategies

Researchers can add implementations under the model and dataset structures and evaluate them against existing benchmarks.

Sources: [2][3]

For reproducing papers and evaluating new continual-learning methods

It fits research workflows that want shared datasets, configuration, and metrics instead of a different codebase per paper.

Sources: [2]

Verify compute, optional dependencies, and reproducibility conditions per method

Methods can require extra packages such as timm, GPU memory, or checkpoints. Use model-specific configs and reproducibility notes for fair comparisons.

Sources: [2][4]

Official sources

  1. [1]aimagelab/mammoth repository(2026-09-21)
  2. [2]Mammoth README(2026-09-21)
  3. [3]Mammoth documentation(2026-09-21)
  4. [4]Mammoth pyproject.toml(2026-09-21)
  5. [5]Mammoth MIT license(2026-09-21)
Supplemental curator note

Mammoth is a Continual Learning research framework, not a PDF/document tool. `pyproject.toml` reports version 0.2.2 with Python 3.10+ and PyTorch 2.1+. The README lists more than 70 methods and 23 datasets, but individual methods can require extra dependencies, checkpoints, and GPU resources.

Try it in 3 steps

  1. 1

    Clone the repository

    The README notes that Mammoth is not yet available on PyPI, so start from the repository.

    git clone https://github.com/aimagelab/mammoth.git && cd mammoth
  2. 2

    Sync dependencies

    Prepare the Python 3.10+ and PyTorch 2.1+ environment from pyproject.toml.

    uv sync
  3. 3

    Run a DER++ benchmark

    Launch a standard continual-learning experiment; dataset downloads and significant compute may be required.

    uv run python main.py --model derpp --dataset seq-cifar100 --model_config best
Check the official README

Growth

Growth trends · Last 30 days

838 Stars

Trend data is still being collected.

Development activity

Last 90 days · weekly

Commits (last 30 days)
0
Open PRs
0

Development activity is still being collected.

Built with

Categories and tags

Categories

Tags

GitHub data

GitHub dataView detailed GitHub data
View on GitHub

GitHub Topics

  • pytorch
  • continual-learning
  • neurips2020
  • knowledge-distillation
  • deep-learning
  • der
  • experience-replay
  • dark-experience-replay
Stars
838
Forks
157
Watchers
12
Open issues
1
Contributors
11
Primary language
Python
License
MIT
Repository last updated
May 20, 2026
Write a related article

Share a guide or use case for this OSS in Markdown. Articles are published after administrator approval.

Report incorrect information

Tell us if any listing information is incorrect or outdated.