Implementation:Hpcaitech ColossalAI ColossalQA Logger

Knowledge Sources	Hpcaitech_ColossalAI
Domains	Logging, Utilities, NLP
Last Updated	2026-02-09 00:00 GMT

Overview

ColossalQALogger is a singleton-pattern logging class that provides verbosity-controlled logging with info, warning, debug, and error levels for the ColossalQA application.

Description

The ColossalQALogger class wraps Python's standard logging module and enforces a singleton pattern per logger name using a private class-level dictionary. Each logging method (except error) accepts a verbose flag that determines whether the message is actually emitted, providing fine-grained control over debug output. A convenience function get_logger is provided to retrieve or create logger instances by name, defaulting to the name "colossalqa".

Usage

Use ColossalQALogger (via the get_logger function) throughout ColossalQA components when you need controlled logging output. Set verbose=True on individual log calls to enable output during debugging, while keeping production logs clean by default.

Code Reference

Source Location

• Repository: Hpcaitech_ColossalAI
• File: applications/ColossalQA/colossalqa/mylogging.py
• Lines: 1-93

Signature

class ColossalQALogger:
    def __init__(self, name):
        ...

    @staticmethod
    def get_instance(name: str):
        ...

    def info(self, message: str, verbose: bool = False) -> None:
        ...

    def warning(self, message: str, verbose: bool = False) -> None:
        ...

    def debug(self, message: str, verbose: bool = False) -> None:
        ...

    def error(self, message: str) -> None:
        ...

def get_logger(name: str = None, level=logging.INFO) -> ColossalQALogger:
    ...

Import

from colossalqa.mylogging import get_logger, ColossalQALogger

I/O Contract

Inputs

Name	Type	Required	Description
name	str	Yes (for __init__)	The name for the logger instance; must be unique across all instances
message	str	Yes	The log message string to emit
verbose	bool	No	Controls whether info, warning, and debug messages are actually logged (default: False)

Outputs

Name	Type	Description
get_instance return	ColossalQALogger	The singleton logger instance for the given name
get_logger return	ColossalQALogger	A logger instance, creating one if it does not exist (default name: "colossalqa")

Usage Examples

from colossalqa.mylogging import get_logger

# Get the default ColossalQA logger
logger = get_logger()

# Log with verbose control
logger.info("Loading data...", verbose=True)   # Will print
logger.info("Loading data...", verbose=False)  # Will not print

# Log warnings and debug messages
logger.warning("Deprecated feature used", verbose=True)
logger.debug("Variable x = 42", verbose=True)

# Error messages are always logged regardless of verbose flag
logger.error("Failed to connect to database")

# Get a named logger
custom_logger = get_logger(name="my_module")
custom_logger.info("Module initialized", verbose=True)

Related Pages

• Environment:Hpcaitech_ColossalAI_ColossalQA_RAG_Environment