diff --git a/webui.py b/webui.py index 671a1b6f6..606834001 100644 --- a/webui.py +++ b/webui.py @@ -456,9 +456,7 @@ def process_images(outpath, func_init, func_sample, prompt, seed, sampler_name, prompts = all_prompts[n * batch_size:(n + 1) * batch_size] seeds = all_seeds[n * batch_size:(n + 1) * batch_size] - uc = None - if cfg_scale != 1.0: - uc = model.get_learned_conditioning(len(prompts) * [""]) + uc = model.get_learned_conditioning(len(prompts) * [""]) if isinstance(prompts, tuple): prompts = list(prompts) @@ -558,9 +556,8 @@ def txt2img(prompt: str, ddim_steps: int, sampler_name: str, toggles: list, ddim err = False seed = seed_to_int(seed) - # print('toggles:', toggles) prompt_matrix = 0 in toggles - normalize_prompt_weights = toggles[1] + normalize_prompt_weights = 1 in toggles skip_save = 2 not in toggles skip_grid = 3 not in toggles use_GFPGAN = 4 in toggles @@ -718,7 +715,7 @@ def img2img(prompt: str, init_info, mask_mode, ddim_steps: int, sampler_name: st seed = seed_to_int(seed) prompt_matrix = 0 in toggles - normalize_prompt_weights = toggles[1] + normalize_prompt_weights = 1 in toggles loopback = 2 in toggles skip_save = 3 not in toggles skip_grid = 4 not in toggles