Files

T

Martin Rizzo f39fbb5f7e feat: Simplify —upscale parameter to a boolean toggle

This commit refactors the handling of the `--upscale` parameter.
Previously, it accepted a float representing the upscaling factor,
which could be confusing for users. Now, `--upscale` functions as
a simple boolean toggle, enabling or disabling the upscaler.

This change allows users to control the upscaling factor directly
within the upscaler node itself.

2025-04-13 10:48:35 -03:00

5.7 KiB

Raw Permalink Blame History

GenParams

GenParams is a dictionary containing parameters used during the generation process, along with supplementary information.

Important Notes

All parameters within GenParams are optional.
The code should gracefully handle cases where parameters are missing by:

Assigning default values.
Performing additional calculations to predict the missing parameter's value.

Parameters

File Parameters

file.name (str): The name of the file containing the main model parameters (only the name and extension, without path)
file.date (str): The date when the file was generated, in ISO 8601 format (e.g., YYYY-MM-DD).

Model Parameters

modelspec.architecture (str): The name of the architecture used by the model.
modelspec.title (str): The title of the model, used for identification.
modelspec.date (str): The date the model was created, in ISO 8601 format (e.g., YYYY-MM-DD).
modelspec.license (str): The license under which the model is distributed.
modelspec.resolution (str): The intended resolution for the model, formatted as <width>x<height> (e.g., "10240x1024").

User-Provided Parameters

user.style (str): The name of the style selected by the user in the UI.
user.prompt (str): The prompt entered by the user in the UI.
user.negative (str): The negative prompt entered by the user in the UI.

Image Parameters

image.scale (float): The scaling factor applied to the image size, based in the original resolution of the model (e.g., 1.22).
image.orientation (str): The orientation of the image (e.g., "landscape", "portrait").
image.aspect_ratio (str): The aspect ratio of the image (e.g., "16:9").
image.enable_upscaler (bool): A flag indicating whether upscaling is enabled (default: False)
image.batch_size (int): The number of images that are generated in each batch (e.g., 1).

Denoising Parameters

denoising.base.prompt (str): The prompt used for the base model.
denoising.base.negative (str): The negative prompt used for the base model.
denoising.base.steps (int): The total number of steps for the base model.
denoising.base.steps_start (int): The step at which the base model's sampler begins the diffusion process.
denoising.base.steps_end (int): The step at which the base model's sampler terminates the diffusion process.
denoising.base.steps_nfactor (int): A factor used to expand or reduce the number of steps for the base model.
denoising.base.cfg (float): The guidance scale used by the base model's sampler.
denoising.base.noise_seed (int): The seed used for generating initial noise used by the base model.
denoising.base.sampler (str): The name of the sampler used by the base model.
denoising.base.scheduler (str): The name of the scheduler used by the base model.

Denoising Parameters (Refiner)

denoising.refiner.prompt (str): The prompt used for the refiner model.
denoising.refiner.negative (str): The negative prompt used for the refiner model.
denoising.refiner.steps (int): The total number of steps for the refiner model.
denoising.refiner.steps_start (int): The step at which the refiner model's sampler begins the diffusion process.
denoising.refiner.steps_end (int): The step at which the refiner model's sampler terminates the diffusion process.
denoising.refiner.steps_nfactor (int): A factor used to expand or reduce the number of steps for the refiner model.
denoising.refiner.cfg (float): The guidance scale used by the refiner model's sampler.
denoising.refiner.noise_seed (int): The seed used for generating initial noise used by the refiner model.
denoising.refiner.sampler (str): The name of the sampler used by the refiner model.
denoising.refiner.scheduler (str): The name of the scheduler used by the refiner model.

Denoising Parameters (Upscaler)

denoising.upscaler.prompt (str): The prompt used for the model during upscaling.
denoising.upscaler.negative (str): The negative prompt used for the model during upscaling.
denoising.upscaler.steps (int): The total number of denoising steps during upscaling.
denoising.upscaler.steps_start (int): The step at which starting denoising for upscaling.
denoising.upscaler.steps_end (int): The step at which ending denoising for upscaling.
denoising.upscaler.steps_nfactor (int): A factor used to expand or reduce the number of denoising steps.
denoising.upscaler.cfg (float): The guidance scale used during upscaling.
denoising.upscaler.noise_seed (int): The seed used for generating initial noise during upscaling.
denoising.upscaler.sampler (str): The name of the sampler used during upscaling.
denoising.upscaler.scheduler (str): The name of the scheduler used during upscaling.

Style Parameters

styles.<name>.base.*:
- styles.<name>.base.prompt (str): The base prompt for a specific style named <name>.
- styles.<name>.base.negative (str): The base negative prompt for a specific style named <name>.
- ...
styles.<name>.refiner.*:
- styles.<name>.refiner.prompt (str): The refiner prompt for a specific style named <name>.
- styles.<name>.refiner.negative (str): The refiner negative prompt for a specific style named <name>.
- ... (Indicates there may be additional refiner style-related parameters)

Final Notes

GenParams is used to pass information between different parts of the code during the generation process. It contains both user-provided and automatically calculated values.
As a dictionary, it can store any parameter that needs to be passed around, but the parameters mentioned here are the most important used in the project.

5.7 KiB Raw Permalink Blame History