Installation

Requires Python 3.10+. We recommend uv.

uvpip

uv pip install strands-transformers   # from PyPI
# from source:
uv pip install -e .
# optional extras:
uv pip install -e ".[audio]"      # soundfile, librosa  (mp3/flac/ogg decode)
uv pip install -e ".[vision]"     # pillow, torchvision (vision models!), opencv, av
uv pip install -e ".[training]"   # trl, peft, accelerate
uv pip install -e ".[docs]"       # mkdocs-material, mkdocstrings
uv pip install -e ".[all]"        # everything

pip install strands-transformers      # from PyPI
# from source:
pip install -e .
pip install -e ".[audio]"
pip install -e ".[vision]"
pip install -e ".[training]"
pip install -e ".[all]"

Verify your install

Fast, no big downloads - 18 real checks:

PYTHONPATH=. python examples/smoke.py     # → "18/18 checks passed"

Optional extras

Extra	Pulls in	Needed for
audio	soundfile, librosa	mp3/flac/ogg decode (WAV works without it)
vision	torchvision, opencv, av	vision-language models + video
training	trl, peft, accelerate	fine-tuning workflows
docs	mkdocs-material, mkdocstrings	building this site

Device selection

device="auto" (the default) picks cuda → mps → cpu, and uses bf16 on GPU.