ML Observability in your notebook
Project description
phoenix.ai
Phoenix enables you to get to MLOps insights at lightning speed. Phoenix focuses on surfacing areas that require critical attention and offers zero-config observability for model drift, performance, and data quality.
_ NOTE: Phoenix is under active development. APIs may change at any time _
Getting started
pip install phoenix
Troubleshooting
If you are using an Apple M1 machine and encounter the error Could not find a local HDF5 installation
, take the following steps:
- Install HDF5 with
arch -arm64 brew install hdf5
. - Find the path to your HDF5 installation with
brew info hdf5
. - Set an environment variable with
export HDF5_DIR=/path/to/your/hdf5/installation
. - Retry the command you ran when you encountered the error.
If are you using an Apple M1 machine and encounter the error incompatible architecture (have (x86_64), need (arm64e))
during installation (for example, while installing hdbscan
), take the following steps:
- Run
softwareupdate --install-rosetta
to install Rosetta2 in order to emulate Intel CPUs on your ARM machine. - Purge the
pip
cache withpip cache purge
. - Retry
pip install -e .
from the repo base directory.
Unstructured data
Phoenix takes advantage of UMAP and HDBSCAN to highlight segments of your data that are areas of critical drift.
Structured / tabular data
Phoenix surfaces up problematic features of your model with regards to drift, performance, and data quality.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for arize_phoenix-0.0.1rc0-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 488cbe78a286f58eef89d2db4493f9c5f32ad474f9061de4faadad7741ba8813 |
|
MD5 | 68f986f27a3eb7ef64f6e4ee2abb5a8a |
|
BLAKE2b-256 | ca88644f5ef887b33516671f0013e326b3909ef1f3fae2f8c7f62d7fe140b6f3 |