search

angelolab / Nimbus

Other
12 stars 1 forks source link

Add `conda` environment configurations for Linux w/ CUDA and macOS w/ Metal #70

Closed srivarra closed 5 months ago

srivarra commented 11 months ago

Is your feature request related to a problem? Please describe. Because tensorflow is pinned to 2.8.0, it is unresolvable for Apple Silicon macs. The tensorflow package for macOS on PyPI is under the name tensorflow-macos for 2.12 and earlier. If it's possible to unpin tensorflow and use anything later than 2.13 , then there isn't a need for tensorflow-macos , and pip install tensorflow would work. For example when I run pip install -e ., I get the following error

ERROR: Could not find a version that satisfies the requirement tensorflow==2.8 (from cell-classification) (from versions: 2.13.0rc0, 2.13.0rc1, 2.13.0rc2, 2.13.0, 2.14.0rc0, 2.14.0rc1)
ERROR: No matching distribution found for tensorflow==2.8

Describe the solution you'd like We should include various conda environment configuration files, one for CUDA on Linux / Windows, and one for MPS/metal (macOS gpu compute).

Describe alternatives you've considered Stop using google products.

Additional context N/A

srivarra commented 5 months ago

Not necessary because of the Nimbus Inference repo.