Linux | conda | Make package/environment creation much faster

• Conda/Micromamba

---

How to speed up the micromamba install process (both installing micromamba itself and, more importantly, creating environments + installing packages with it).

1. Install micromamba as quickly as possible

The fastest ways to get micromamba on your system:

• Recommended (one-liner, works on Linux/macOS/Git Bash on Windows):

  "${SHELL}" <(curl -L micro.mamba.pm/install.sh)

  This is the official quickest method.

• macOS with Homebrew (very fast if you already have brew):

  brew install micromamba

• Manual download (fastest for scripts/CI, no shell init needed initially):

  # Linux x86_64 example
  curl -Ls https://micro.mamba.pm/api/micromamba/linux-64/latest | tar -xvj bin/micromamba
  ./bin/micromamba shell init -s bash -r ~/micromamba   # or your preferred prefix
  source ~/.bashrc

After installation, run micromamba self-update to get the latest version (newer releases often include solver and download improvements).

Set a custom root prefix early if you want (e.g., on a fast SSD or large disk):

export MAMBA_ROOT_PREFIX=~/micromamba   # or /fast/ssd/micromamba

2. Make package/environment creation much faster

Micromamba (and mamba) is already dramatically faster than classic conda thanks to its C++ solver (libsolv) and parallel operations. Here are the key optimizations:

• Use only conda-forge (biggest single speedup):

  micromamba config append channels conda-forge
  micromamba config append channels nodefaults   # disable defaults channel
  micromamba config set channel_priority strict

  Mixing channels (especially defaults + conda-forge) slows down the solver a lot.

• Create environments from a YAML file (instead of many separate install commands):

  micromamba create -n myenv -f environment.yml -y

  Or inline:

  micromamba create -n myenv python=3.12 numpy pandas -c conda-forge -y

• Enable parallel downloads and fast settings: Micromamba already uses multi-threading for downloads and solving by default. You can tune it with these config options (add to ~/.condarc or ~/.mambarc):

  # In ~/.condarc
  solver: libmamba          # usually default
  channel_priority: strict
  # Optional performance tweaks:
  # fetch_threads: 0        # 0 = auto (uses all cores; try higher/lower if network is bottleneck)
  # repodata_threads: 0     # auto for repodata fetching

• Leverage the shared package cache: Micromamba caches downloaded packages and repodata (channel indexes) in $MAMBA_ROOT_PREFIX/pkgs/.
  • First install is slower (downloads everything).
    
  • Subsequent installs are much faster because it reuses hard links from the cache.
    Keep the same MAMBA_ROOT_PREFIX across projects for maximum reuse.

• Offline / pre-download mode (great for repeated setups or air-gapped machines):

  # On a machine with internet
  micromamba create -n myenv --download-only -f environment.yml
  
  # Then copy the pkgs cache or use --offline on target machine
  micromamba create -n myenv --offline -f environment.yml

• Other practical tips:
  • Avoid running installs while already inside an activated environment (can cause repeated activation overhead in some cases).
  • For CI/Docker: Use the official mambaorg/micromamba Docker image or the setup-micromamba GitHub Action — it installs in ~1 second.
  • On slow networks: The bottleneck is often downloading repodata/packages. Use a local mirror or proxy if possible (advanced).
  • Clean up occasionally: micromamba clean --all -y to free space (but keep cache for speed).