One more, possible to also change the "sampling method" with sets?

[magicbot42]

With sdxl, I find myself using heun sampler for initial image, but using RealisticVision with "DPM++ 2M Karras" for img2img refinement for example sometimes. Would be nice to have these in the script, and not have to set these each time.

[ThereforeGames]

Hi @magicbot42,

Yep, you want sampler_name for that. Example:

[sets sampler_name="DPM++ 2M Karras"]

[magicbot42]

Thank you! @ThereforeGames By chance is there also an option for the "upscaler method", and whether hires_fix is toggled?

[magicbot42]

Or is there somewhere I can see a list of what options are avail?

[ThereforeGames]

The variable names are derived from the WebUI's p object. For a time, A1111 was frequently updating/changing some of these variables around so I haven't documented them in the manual, but perhaps I should soon.

Here's the full list:

['__class__', '__delattr__', '__dict__', '__dir__', '__doc__', '__eq__', '__format__', '__ge__', '__getattribute__', '__gt__', '__hash__', '__init__', '__init_subclass__', '__le__', '__lt__', '__module__', '__ne__', '__new__', '__reduce__', '__reduce_ex__', '__repr__', '__setattr__', '__sizeof__', '__str__', '__subclasshook__', '__weakref__', 'all_hr_negative_prompts', 'all_hr_prompts', 'all_negative_prompts', 'all_prompts', 'all_seeds', 'all_subseeds', 'applied_old_hires_behavior_to', 'batch_size', 'c', 'cached_c', 'cached_hr_c', 'cached_hr_uc', 'cached_uc', 'calculate_hr_conds', 'cfg_scale', 'close', 'color_corrections', 'ddim_discretize', 'denoising_strength', 'depth2img_image_conditioning', 'disable_extra_networks', 'do_not_reload_embeddings', 'do_not_save_grid', 'do_not_save_samples', 'edit_image_conditioning', 'enable_hr', 'eta', 'extra_generation_params', 'extra_network_data', 'get_conds_with_caching', 'get_token_merging_ratio', 'height', 'hr_c', 'hr_extra_network_data', 'hr_negative_prompt', 'hr_negative_prompts', 'hr_prompt', 'hr_prompts', 'hr_resize_x', 'hr_resize_y', 'hr_sampler_name', 'hr_scale', 'hr_second_pass_steps', 'hr_uc', 'hr_upscale_to_x', 'hr_upscale_to_y', 'hr_upscaler', 'img2img_image_conditioning', 'init', 'inpainting_image_conditioning', 'is_hr_pass', 'is_using_inpainting_conditioning', 'iteration', 'n_iter', 'negative_prompt', 'negative_prompts', 'outpath_grids', 'outpath_samples', 'overlay_images', 'override_settings', 'override_settings_restore_afterwards', 'parse_extra_network_prompts', 'paste_to', 'prompt', 'prompt_for_display', 'prompts', 'restore_faces', 's_churn', 's_min_uncond', 's_noise', 's_tmax', 's_tmin', 'sample', 'sampler', 'sampler_name', 'sampler_noise_scheduler_override', 'script_args', 'scripts', 'sd_model', 'seed', 'seed_resize_from_h', 'seed_resize_from_w', 'seeds', 'setup_conds', 'setup_prompts', 'step_multiplier', 'steps', 'styles', 'subseed', 'subseed_strength', 'subseeds', 'tiling', 'token_merging_ratio', 'token_merging_ratio_hr', 'truncate_x', 'truncate_y', 'txt2img_image_conditioning', 'uc', 'unclip_image_conditioning', 'user', 'width']

You can probably use enable_hr and hr_upscaler but you may need to supply some of the related variables as well, such as hr_sampler_name - I'm not sure what these variables default to. If you have any trouble getting it to work, let me know and I can run some tests.

[magicbot42]

Amazing. Thanks. Will try it out.

@ThereforeGames Sorry for all the questions but one more. Is there a way to multiply numbers? For example, sdxl takes much higher weights than work on other models so I'd like to multiple the weight variables i have created down by some factor if "sd_base" = "sd1".

[ThereforeGames]

Sure, you can use [eval] for arithmetic, e.g. [eval]0.5 * 3[/eval] returns 1.5.

You can also insert arithmetic directly into the arguments of some shortcodes, including [sets]. So maybe you want something like this:

[if sd_base="sd1"][sets my_var="{get some_var} * 0.5"][/if]

In general, I recommend trying small snippets of code in the "Dry Run" box to get a better sense of the syntax before building a larger template for Stable Diffusion tasks.

[magicbot42]

Thanks again. This tool keeps getting better, the more I use it!