v1.9.0
Open-CE Version 1.9.0
This is release 1.9.0 of Open Cognitive Environment (Open-CE).
Package Versions
A release of Open-CE consists of the environment files within the open-ce repository and a collection of feedstock repositories. The feedstock repositories contain recipes for various python packages. The following packages (among others) are part of this release:
| Package | Version |
|---|---|
| dali | 1.25.0 |
| deepspeed | 0.8 |
| liblightgbm | 3.3.2 and 3.3.5 |
| av | 10.0 |
| bazel | 5.3.0 |
| boost_mp11 | 1.76.0 |
| cmdstan | 2.31.0 |
| cmdstanpy | 1.1.0 |
| cudatoolkit | 11.8.0 |
| cudatoolkit-dev | 11.8.0 |
| cudnn | 8.8.1 |
| datasets | 2.10.1 |
| dm-tree | 0.1.7 |
| grpc-cpp | 1.41.0 |
| gtest | 1.13.0 |
| horovod | 0.28.0 |
| jpeg-turbo | 2.1.4 |
| keras | 2.12.0 |
| libdate | 3.0.1 |
| libflac | 1.3.3 |
| libiconv | 1.16 |
| libsndfile | 1.0.31 |
| libsolv | 0.7.19 |
| magma | 2.6.1 |
| mamba | 1.4.1 |
| libmamba | 1.4.1 |
| nccl | 2.17.1 |
| numactl | 2.0.16 |
| onnx | 1.13.1 |
| onnxconverter-common | 1.13.0 |
| onnxmltools | 1.11.2 |
| onnxruntime | 1.14.1 |
| libopencv | 4.7.0 |
| openblas | 0.3.23 |
| openmpi | 4.1.4 |
| optional-lite | 3.4.0 |
| orbit-ml | 1.1.4.2 |
| orc | 1.8.2 |
| arrow-cpp | 11.0.0 |
| prophet | 1.1.2 |
| pytorch-base | 2.0.0 |
| pytorch-lightning | 2.0.1 |
| pytorch_geometric | 2.3.0 |
| pytorch-lightning-bolts | 0.6.0 |
| pytorch_scatter | 2.1.1 |
| pytorch_sparse | 0.6.17 |
| torchmetrics | 0.11.4 |
| safeint | 3.0.26 |
| sentencepiece | 0.1.97 |
| skl2onnx | 1.14 |
| spdlog | 1.9.2 |
| tensorboard-data-server | 0.7.0 |
| tensorboard | 2.12.2 |
| tensorflow-addons | 0.19.0 for cpu only |
| tensorflow-datasets | 4.9.2 |
| tensorflow-estimator | 2.12.0 |
| tensorflow-base | 2.12.0 |
| tensorflow-hub | 0.13.0 |
| tensorflow-io | 0.32.0 |
| tensorflow-io-gcs-filesystem | 0.32.0 |
| tensorflow-metadata | 1.13.1 |
| tensorflow-model-optimization | 0.7.4 |
| tensorflow-probability | 0.19.0 |
| tensorflow-text | 2.12.0 |
| tf2onnx | 1.13.0 |
| torchtext-base | 0.15.1 |
| torchvision-base | 0.15.1 |
| libxgboost | 1.7.5 and 1.6.2 |
| xgboost | 1.7.5 and 1.6.2 |
| apache-beam | 2.46.0 |
| Jax | 0.4.7 |
| Array-record | 0.2.0 |
| bazel-toolchain | 0.1.5 |
| ffmpeg | 4.2.2 |
| Protobuf | 4.21.12 |
| libprotobuf | 3.21.12 |
| lightning-app | 2.0.1 |
| lightning-cloud | 0.5.32 |
| lightning-fabric | 2.0.1 |
| lightning-utilities | 0.8.0 |
This release of Open-CE supports NVIDIA's CUDA version 11.8 well as Python 3.9 and 3.10. Support for CUDA 11.2 and python 3.8 will be added back in Open-CE 1.9.1. CUDA 11.4 support is dropped.
Getting Started
To get started with this release, see the main readme
Known Issues
There is an issue in isolating CPU packages from CUDA packages when built on same channel. Hence it is recommended to build CPU and CUDA packages into separate channels.
The following pip check failures are known:
deepspeed 0.7.5+06e00f61 requires ninja, which is not installed.
multiprocess 0.70.12.2 has requirement dill>=0.3.4, but you have dill 0.3.1.1.
lightning-bolts 0.6.0.post1 has requirement lightning-utilities==0.7.1, but you have lightning-utilities 0.8.0.
tf-model-optimization-nightly 0.7.4.dev0 has requirement absl-py~=1.2, but you have absl-py 1.0.0.