Bug Description

There is an inconsistency between the Python version classifiers and the keras dependency version in pyproject.toml. The dependency specifies keras>=3.0.0, but keras 3.x requires Python >=3.11, while the classifiers and requires-python in AutoKeras allow Python 3.8 to 3.11. As a result, users with Python <3.11 may encounter install or runtime issues.

Relevant snippets:

"keras>=3.0.0",

classifiers = [
    "Programming Language :: Python :: 3.8",
    "Programming Language :: Python :: 3.9",
    "Programming Language :: Python :: 3.10",
    "Programming Language :: Python :: 3.11",
]
requires-python = ">=3.8"

Bug Reproduction

• Set up a Python 3.8 / 3.9 / 3.10 environment
• Install autokeras per instruction
• Observe installation or usage errors due to keras>=3.0.0 requiring Python >=3.11

Expected Behavior

• The dependency and classifiers should be consistent, and supported Python versions should match the requirements of keras>=3.0.0
• Users with Python <3.11 should not be led to believe they can install/fully use autokeras if keras 3.0+ is required

Setup Details

• OS type and version: any
• Python: 3.8, 3.9, 3.10, 3.11
• autokeras: latest
• keras-tuner: any
• scikit-learn: any
• numpy: any
• pandas: any
• tensorflow: any

Additional context

The dependency and classifiers should be aligned. Please update the supported Python versions or the Keras dependency range to avoid confusion and installation problems.

Solution

Refine the Python Classifier, similar to Keras, to be consistent in method.