Installation#
Requirements#
Python ≥ 3.10
PyTorch (the only runtime dependency)
Install#
Packages are published to Embedl’s own package index at https://deploy.embedl.com/downloads/ rather than to PyPI. The index requires an Embedl Hub account: create one at https://hub.embedl.com and generate an API key on your profile page (https://hub.embedl.com/profile). Pass the key as the password in the index URL; the username is ignored. The recommended install includes the TensorRT backend and the mixed-precision search tooling:
pip install --extra-index-url "https://hub:<API_KEY>@deploy.embedl.com/downloads/" "embedl-deploy[tensorrt,search]"
--extra-index-url keeps PyPI as the source for dependencies such as
PyTorch and Optuna. If you prefer not to put the key on the command line,
set it once in pip’s configuration instead:
pip config set global.extra-index-url "https://hub:<API_KEY>@deploy.embedl.com/downloads/"
pip install "embedl-deploy[tensorrt,search]"
Browse https://deploy.embedl.com/downloads/ in a browser (username hub,
password your API key) to see which versions are available.
Extras#
The base package alone (pip install embedl-deploy) is enough to build
models in PyTorch and export to ONNX. The extras add:
Extra |
Adds |
Target |
|---|---|---|
|
the |
NVIDIA GPUs |
|
Optuna, for mixed-precision search (see the tutorials) |
any |
See Backends for how the active backend is selected at runtime.
For model compilation, you may need additional tools such as TensorRT
(trtexec), but these are not required to create deployment-ready models
in PyTorch and export them to ONNX.