Public/automatic
1
0
Fork 0
mirror of https://github.com/vladmandic/automatic synced 2026-09-19 01:04:32 +02:00
Code Issues Packages Projects Releases Wiki Activity

prompt error handling

Browse Source 
Signed-off-by: Vladimir Mandic <mandic00@live.com>
This commit is contained in:
Vladimir Mandic
2025-07-25 12:38:25 -04:00
parent fd2dde9154
commit 0b8001c0fa
3 changed files with 27 additions and 19 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/processing_args.py
+24 -17
View File
@@ -157,7 +157,7 @@ def set_pipeline_args(p, model, prompts:list, negative_prompts:list, prompts_2:t
'StableCascade' in model.__class__.__name__ or
'Flux' in model.__class__.__name__ or
'Chroma' in model.__class__.__name__ or
'HiDreamImagePipeline' in model.__class__.__name__ # hidream-e1 has different embeds
'HiDreamImagePipeline' in model.__class__.__name__
):
try:
prompt_parser_diffusers.embedder = prompt_parser_diffusers.PromptEmbedder(prompts, negative_prompts, steps, clip_skip, p)
@@ -173,23 +173,28 @@ def set_pipeline_args(p, model, prompts:list, negative_prompts:list, prompts_2:t
if 'prompt' in possible:
if 'OmniGen' in model.__class__.__name__:
prompts = [p.replace('|image|', '<img><|image_1|></img>') for p in prompts]
if 'HiDreamImage' in model.__class__.__name__ and prompt_parser_diffusers.embedder is not None:
if ('HiDreamImage' in model.__class__.__name__) and (prompt_parser_diffusers.embedder is not None):
args['pooled_prompt_embeds'] = prompt_parser_diffusers.embedder('positive_pooleds')
prompt_embeds = prompt_parser_diffusers.embedder('prompt_embeds')
args['prompt_embeds_t5'] = prompt_embeds[0]
args['prompt_embeds_llama3'] = prompt_embeds[1]
elif hasattr(model, 'text_encoder') and hasattr(model, 'tokenizer') and 'prompt_embeds' in possible and prompt_parser_diffusers.embedder is not None:
args['prompt_embeds'] = prompt_parser_diffusers.embedder('prompt_embeds')
if 'StableCascade' in model.__class__.__name__:
args['prompt_embeds_pooled'] = prompt_parser_diffusers.embedder('positive_pooleds').unsqueeze(0)
elif 'XL' in model.__class__.__name__:
args['pooled_prompt_embeds'] = prompt_parser_diffusers.embedder('positive_pooleds')
elif 'StableDiffusion3' in model.__class__.__name__:
args['pooled_prompt_embeds'] = prompt_parser_diffusers.embedder('positive_pooleds')
elif 'Flux' in model.__class__.__name__:
args['pooled_prompt_embeds'] = prompt_parser_diffusers.embedder('positive_pooleds')
elif 'Chroma' in model.__class__.__name__:
args['prompt_attention_mask'] = prompt_parser_diffusers.embedder('prompt_attention_masks')
elif hasattr(model, 'text_encoder') and hasattr(model, 'tokenizer') and ('prompt_embeds' in possible) and (prompt_parser_diffusers.embedder is not None):
embeds = prompt_parser_diffusers.embedder('prompt_embeds')
if embeds is None:
shared.log.warning('Prompt parser encode: empty prompt embeds')
args['prompt'] = prompts
else:
args['prompt_embeds'] = embeds
if 'StableCascade' in model.__class__.__name__:
args['prompt_embeds_pooled'] = prompt_parser_diffusers.embedder('positive_pooleds').unsqueeze(0)
elif 'XL' in model.__class__.__name__:
args['pooled_prompt_embeds'] = prompt_parser_diffusers.embedder('positive_pooleds')
elif 'StableDiffusion3' in model.__class__.__name__:
args['pooled_prompt_embeds'] = prompt_parser_diffusers.embedder('positive_pooleds')
elif 'Flux' in model.__class__.__name__:
args['pooled_prompt_embeds'] = prompt_parser_diffusers.embedder('positive_pooleds')
elif 'Chroma' in model.__class__.__name__:
args['prompt_attention_mask'] = prompt_parser_diffusers.embedder('prompt_attention_masks')
else:
args['prompt'] = prompts
if 'negative_prompt' in possible:
@@ -406,11 +411,13 @@ def set_pipeline_args(p, model, prompts:list, negative_prompts:list, prompts_2:t
clean['generator'] = f'{generator[0].device}:{[g.initial_seed() for g in generator]}'
clean['parser'] = parser
for k, v in clean.copy().items():
if isinstance(v, torch.Tensor) or isinstance(v, np.ndarray):
if v is None:
clean[k] = None
elif isinstance(v, torch.Tensor) or isinstance(v, np.ndarray):
clean[k] = v.shape
if isinstance(v, list) and len(v) > 0 and (isinstance(v[0], torch.Tensor) or isinstance(v[0], np.ndarray)):
elif isinstance(v, list) and len(v) > 0 and (isinstance(v[0], torch.Tensor) or isinstance(v[0], np.ndarray)):
clean[k] = [x.shape for x in v]
if not debug_enabled and k.endswith('_embeds'):
elif not debug_enabled and k.endswith('_embeds'):
del clean[k]
clean['prompt'] = 'embeds'
task = str(sd_models.get_diffusers_task(model)).replace('DiffusersTaskType.', '')