Fix polars for 1.2.1 #13868
You are viewing an older attempt in the history of this workflow run.
View latest attempt.
pr.yaml
on: push
Matrix: devcontainer / build
Matrix: conda-cpp-build / build
Matrix: wheel-build-cudf / build
conda-cpp-checks
/
check-symbols
5m 12s
conda-java-tests
/
build
8m 54s
Matrix: conda-cpp-tests / tests
Matrix: conda-python-build / build
Matrix: pandas-tests / test
Matrix: unit-tests-cudf-pandas / test
Matrix: wheel-build-cudf-polars / build
Matrix: wheel-build-dask-cudf / build
Matrix: wheel-tests-cudf / test
conda-notebook-tests
/
build
13m 38s
docs-build
/
build
13m 23s
Matrix: conda-python-cudf-tests / tests
Matrix: conda-python-other-tests / tests
Matrix: wheel-tests-cudf-polars / test
Matrix: wheel-tests-dask-cudf / test
pr-builder
/
run
Annotations
16 errors
conda-python-cudf-tests / tests (amd64, 3.9, 11.8.0, rockylinux8, v100, earliest)
The self-hosted runner: linux-amd64-gpu-v100-earliest-1-bbxjq-runner-kp852 lost communication with the server. Verify the machine is running and has a healthy network connection. Anything in your workflow that terminates the runner process, starves it for CPU/Memory, or blocks its network access can cause this error.
|
conda-python-other-tests / tests (amd64, 3.11, 12.5.1, ubuntu22.04, v100, latest)
The self-hosted runner: linux-amd64-gpu-v100-latest-1-56xmt-runner-9gcp4 lost communication with the server. Verify the machine is running and has a healthy network connection. Anything in your workflow that terminates the runner process, starves it for CPU/Memory, or blocks its network access can cause this error.
|
conda-notebook-tests / build
The self-hosted runner: linux-amd64-gpu-v100-latest-1-56xmt-runner-zz5dm lost communication with the server. Verify the machine is running and has a healthy network connection. Anything in your workflow that terminates the runner process, starves it for CPU/Memory, or blocks its network access can cause this error.
|
docs-build / build
The self-hosted runner: linux-amd64-gpu-v100-latest-1-56xmt-runner-dl5r4 lost communication with the server. Verify the machine is running and has a healthy network connection. Anything in your workflow that terminates the runner process, starves it for CPU/Memory, or blocks its network access can cause this error.
|
unit-tests-cudf-pandas / 12.5.1, 3.9, amd64, ubuntu22.04, v100
The run was canceled by @lithomas1.
|
conda-python-other-tests / tests (arm64, 3.10, 12.0.1, ubuntu20.04, a100, latest)
The run was canceled by @lithomas1.
|
pandas-tests / 12.5.1, 3.9, amd64, ubuntu22.04, v100
The run was canceled by @lithomas1.
|
conda-cpp-tests / tests (arm64, 3.10, 12.0.1, ubuntu20.04, a100, latest)
The run was canceled by @lithomas1.
|
conda-python-cudf-tests / tests (arm64, 3.10, 12.0.1, ubuntu20.04, a100, latest)
The run was canceled by @lithomas1.
|
wheel-tests-cudf / 11.8.0, 3.11, arm64, ubuntu20.04, a100
The run was canceled by @lithomas1.
|
wheel-tests-cudf-polars / 12.5.1, 3.9, amd64, ubuntu22.04, v100
The run was canceled by @lithomas1.
|
wheel-tests-dask-cudf / 12.5.1, 3.9, amd64, ubuntu22.04, v100
The run was canceled by @lithomas1.
|
wheel-tests-cudf / 12.5.1, 3.9, amd64, ubuntu22.04, v100
The run was canceled by @lithomas1.
|
conda-python-cudf-tests / tests (amd64, 3.11, 12.5.1, ubuntu22.04, v100, latest)
The run was canceled by @lithomas1.
|
conda-python-cudf-tests / tests (amd64, 3.11, 12.5.1, ubuntu22.04, v100, latest)
Value cannot be null. (Parameter 'ContainerId')
|
conda-python-cudf-tests / tests (amd64, 3.11, 12.5.1, ubuntu22.04, v100, latest)
The operation was canceled.
|