RAPIDS Logger

0.2.3 · active · verified Thu Apr 16

rapids-logger is a core logging framework within the RAPIDS ecosystem, built around the high-performance C++ `spdlog` library. Its primary goal is to provide project-specific loggers easily, ensuring that custom logging implementations do not leak `spdlog` or `fmt` symbols, thus allowing safe coexistence of different RAPIDS projects in the same environment. While it has Python bindings for packaging, its direct Python API for general-purpose logging is not extensively documented; it primarily serves as the underlying logging mechanism for other RAPIDS Python libraries like `cuML`. The current version is 0.2.3, and releases align with the broader RAPIDS CalVer (YY.MM.PP) versioning scheme for patch releases.

Common errors

```
ModuleNotFoundError: No module named 'rapids_logger'
```
cause The `rapids-logger` Python package is not installed in the current environment or the Python environment is not correctly activated.

fix Install the package using `pip install rapids-logger` or `conda install -c rapidsai -c conda-forge rapids-logger`. Ensure your Python environment is correctly activated before running your script.
```
TypeError: 'module' object is not callable (when trying to call rapids_logger directly as a logging function)
```
cause Attempting to use `rapids_logger` as a direct logging function or to instantiate a logger directly, which is not its exposed Python API.

fix Understand that `rapids-logger` primarily serves as an internal dependency for other RAPIDS Python libraries. Configure logging via the specific RAPIDS library's API or environment variables, rather than trying to directly instantiate a logger from the `rapids_logger` module in Python. Refer to the documentation of the specific RAPIDS library you are using for its logging configuration options.

Warnings

gotcha rapids-logger is fundamentally a C++ library. While Python bindings and wheels are provided, direct high-level Python API for creating and managing loggers (like Python's built-in `logging` module or `loguru`) is not a primary documented feature. Its Python package primarily facilitates its underlying use within other RAPIDS Python libraries.
Fix: For Python applications, primarily configure logging via the specific RAPIDS library you are using (e.g., `cuml`, `cudf`) or through environment variables if supported by the upstream RAPIDS library. Do not expect `rapids-logger` to expose a generic Python logging interface.
breaking For C++ projects building within the RAPIDS ecosystem, `rapids-logger` has replaced direct dependencies on `spdlog` and `fmt`. Projects that previously linked directly against `spdlog` or `fmt` for logging may need to update their CMake configurations to use `rapids-logger` instead.
Fix: Update CMakeLists.txt to remove direct `spdlog` and `fmt` dependencies where `rapids-logger` is intended as the logging provider, and integrate `rapids-logger` via `rapids_make_logger` or similar RAPIDS CMake functions. This mainly impacts developers building RAPIDS components from source or extending them in C++.
gotcha There is an unrelated CI tool also named 'rapids-logger' which is part of the `gha-tools` repository. Ensure you are referencing the correct `rapidsai/rapids-logger` for the logging framework.
Fix: Verify the source repository (github.com/rapidsai/rapids-logger) and PyPI/Conda package (`rapids-logger` from `rapidsai` channel) to confirm you are using the correct library.

Install

pip install rapids-logger Stable release
conda install -c rapidsai -c conda-forge rapids-logger Conda installation (recommended for RAPIDS)

Imports

Logger
wrong:
```
from rapids_logger import Logger
```
correct:
```
N/A (Typically used indirectly)
```
rapids-logger is primarily a C++ library. While Python wheels exist, a high-level Python API for direct logger instantiation and use, analogous to Python's `logging` module, is not a primary documented feature. Logging is typically configured through higher-level RAPIDS Python libraries that use `rapids-logger` internally.

Quickstart

Due to `rapids-logger` being predominantly a C++ framework, direct Python usage for general logging is not commonly documented. Python users typically configure logging through environment variables or specific API calls provided by higher-level RAPIDS Python libraries (e.g., `cuml.accel.install(log_level='debug')`). The provided quickstart demonstrates checking the installed version and notes the indirect usage pattern.

# rapids-logger is primarily a C++ library and its direct Python API for general-purpose logging
# is not a primary documented feature. Python users typically interact with logging
# through other RAPIDS Python libraries (e.g., cuML) that use rapids-logger internally.
#
# Example of configuring logging in cuML (which uses rapids-logger implicitly):
# import cuml
# cuml.accel.install(log_level='debug')
#
# To see if rapids-logger's C++ components are correctly installed and available via Python bindings:
import rapids_logger
print(f"rapids_logger version: {rapids_logger.__version__}")
# Further direct interaction would depend on specific, often undocumented, Python bindings.

view raw JSON →