Description of Evaluation Configuration Parameters

• model_name:
  Type: str
  Description: Name of the model to use. Options include "SD-v1-4", "SD-V2", "SD-V2-1", etc. required: False

• encoder_model_name_or_path
  Description: Model name or path for the encoder. Type: str
  Example: CompVis/stable-diffusion-v1-4

• target_ckpt:
  Type: str
  Description: Path to the target checkpoint.

• If empty, the script will load the default model weights.

• If provided, it supports both Diffusers-format checkpoints (directory) and CompVis checkpoints (file ending with .pt). For CompVis, use the checkpoint of the model saved as Diffuser format.

• save_path:
  Type: str
  Description: Directory where the generated images will be saved.

• prompts_path:
  Type: str
  Description: Path to the CSV file containing prompts, evaluation seeds, and case numbers.
  Default: "data/prompts/visualization_example.csv"

• guidance_scale:
  Type: float
  Description: Parameter that controls the classifier-free guidance during generation.
  Default: 7.5

• image_size:
  Type: int
  Description: Dimensions of the generated images (height and width).
  Default: 512

• ddim_steps:
  Type: int
  Description: Number of denoising steps (used in the diffusion process).
  Default: 100

• num_samples:
  Type: int
  Description: Number of samples generated for each prompt.
  Default: 1

• from_case:
  Type: int
  Description: Minimum case number from which to start generating images.
  Default: 0

• folder_suffix:
  Type: str
  Description: Suffix added to the output folder name for visualizations.

• prompt_path:
  Type: str
  Description: Path to the CSV file containing prompts for evaluation.
  Example: "data/prompts/coco_10k.csv"

• devices:
  Type: str
  Description: Comma-separated list of device IDs to be used during evaluation.
  Example: "0,0"