Implementation:Junyanz Pytorch CycleGAN and pix2pix TestOptions Parse

**Metadata**
Knowledge Sources	pytorch-CycleGAN-and-pix2pix
Domains	Image-to-Image Translation, Configuration Management, Inference Pipeline
Last Updated	2026-02-09

Overview

The TestOptions class in options/test_options.py extends BaseOptions to add inference-specific command-line arguments and override default values for testing. BaseOptions in options/base_options.py provides the shared argument parser and parsing logic.

Description

TestOptions inherits from BaseOptions and overrides the initialize() method to register additional command-line arguments specific to inference. It also modifies default values for existing arguments to better suit the testing workflow.

BaseOptions handles:

Creating an argparse.ArgumentParser with shared arguments
The parse() method that processes command-line arguments, sets up GPU devices, and saves options to disk
print_options() for displaying the current configuration

Usage

Instantiated at the top of test.py to configure the inference pipeline. The parse() method returns an options namespace object that is passed to model and dataset constructors.

Code Reference

Source Location

File	Lines
options/test_options.py	L1-23
options/base_options.py	L9-127

Signature

class TestOptions(BaseOptions):
    """This class includes test options.

    It also includes shared options defined in BaseOptions.
    """

    def initialize(self, parser):
        """Define test-specific parameters.

        Parameters added:
            --results_dir (str)    -- saves results here (default: ./results/)
            --aspect_ratio (float) -- aspect ratio of result images (default: 1.0)
            --phase (str)          -- train, val, test, etc. (default: test)
            --eval                 -- use eval mode during test time
            --num_test (int)       -- how many test images to run (default: 50)

        Default overrides:
            --model  -> 'test'     (single-direction generator only)
            --load_size -> crop_size (no random cropping during test)
        """

class BaseOptions():
    """This class defines options used during both training and test time."""

    def __init__(self, cmd_line=None):
        """Reset the class; used to initialize options."""

    def initialize(self, parser):
        """Define shared options for training and testing."""

    def gather_options(self):
        """Initialize parser with basic options, then add model/dataset options."""

    def print_options(self, opt):
        """Print and save options to a text file."""

    def parse(self):
        """Parse options, create checkpoints dir, and set up GPU devices.
        Returns:
            opt (Namespace) -- parsed options namespace
        """

Import

from options.test_options import TestOptions

opt = TestOptions().parse()

I/O Contract

**Inputs (Command-Line Arguments -- Test-Specific)**
Argument	Type	Default	Description
--results_dir	str	./results/	Directory to save test results
--aspect_ratio	float	1.0	Aspect ratio for output images
--phase	str	test	Phase identifier (determines data subdirectory)
--eval	flag	False	Enable evaluation mode (model.eval())
--num_test	int	50	Maximum number of test images to process

**Inputs (Command-Line Arguments -- Base/Shared)**
Argument	Type	Default	Description
--dataroot	str	(required)	Path to dataset root
--name	str	experiment_name	Experiment name (determines checkpoint/results subdirectory)
--model	str	test (overridden)	Model type: cycle_gan, pix2pix, test, colorization
--input_nc	int	3	Number of input image channels
--output_nc	int	3	Number of output image channels
--ngf	int	64	Number of generator filters
--netG	str	resnet_9blocks	Generator architecture name
--norm	str	instance	Normalization layer type
--no_dropout	flag	False	Disable dropout in generator
--load_size	int	crop_size (overridden)	Scale images to this size
--crop_size	int	256	Crop images to this size
--gpu_ids	str	0	Comma-separated GPU IDs (e.g., 0,1,2; -1 for CPU)

**Outputs**
Output	Type	Description
opt	argparse.Namespace	Parsed options object with all configuration values as attributes

Usage Examples

from options.test_options import TestOptions

# Parse test options from command line
opt = TestOptions().parse()

# Access parsed values
print(opt.dataroot)      # Dataset path
print(opt.name)          # Experiment name
print(opt.results_dir)   # Results output directory
print(opt.num_test)      # Number of test images
print(opt.eval)          # Whether to use eval mode

# Example command line invocation
python test.py --dataroot ./datasets/horse2zebra/testA \
    --name horse2zebra_pretrained \
    --model test \
    --no_dropout \
    --num_test 100 \
    --eval \
    --results_dir ./my_results/

Related Pages

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment