Upgrade RedisAI to 1.2.7 #234
Conversation
Codecov Report
Additional details and impacted files@@ Coverage Diff @@
## develop #234 +/- ##
===========================================
- Coverage 84.31% 84.25% -0.07%
===========================================
Files 56 56
Lines 3284 3284
===========================================
- Hits 2769 2767 -2
- Misses 515 517 +2
|
doc/installation.rst
Outdated
| * - ONNX_ | ||
| - 1.9 | ||
| * - ONNX_ Runtime | ||
| - 1.11.1 |
There was a problem hiding this comment.
If these updates require RAI v1.2.7, we shouldn't claim that we still support older versions. There is some ambiguity here as to whether we're pushing everything forward or not.
There was a problem hiding this comment.
run_tests.yml still tries to test with older versions of RedisAI
There was a problem hiding this comment.
For now, we are keeping 1.2.3, 1.2.5, 1.2.7. I've updated the table to reflect the various versions we support
There was a problem hiding this comment.
Thanks for addressing this. Some feedback that I think needs to be addressed first. It's also unclear to me whether we're trying to support older versions of RedisAI or not -- some of the code does support older versions, but some seems to only support v1.2.7
- Rename torch_gpu_suffix to torch_cuda_suffix - Update documentation to be more specific about RedisAI and ML library dependencies
ashao
force-pushed
the
redisai_upgrade
branch
from
16549b4
to
3613592
Compare
With the release of RedisAI 1.2.7, we can bump the supported versions of pyTorch, Tensorflow, and ONNX to more recent versions. Additionally, while updating the dependencies for RedisAI 1.2.7, the installation of the pyTorch backends with GPU support should now be more robust because we pin a more prescriptive version of
torchandtorchvision.The new versions were tested on an internal HPE machine with on V100s.