Enable A1111 and Full parsing conversion

modules/prompt_parser_diffusers.py

@@ -1,9 +1,25 @@
 import torch
 import modules.shared as shared
+import modules.prompt_parser as prompt_parser
 from compel import Compel, ReturnedEmbeddingsType
 import diffusers
 import typing
 
+def convert_to_compel(prompt: str):
+    if prompt is None:
+      return None
+    all_schedules = prompt_parser.get_learned_conditioning_prompt_schedules(prompt, 100)[0] #100 should be steps, but doesn't actually matter because we can't schedule yet
+    output_list = prompt_parser.parse_prompt_attention(all_schedules[0][1])
+    converted_prompt = []
+    for subprompt, weight in output_list:
+      if subprompt != " ":
+        if weight == 1:
+          converted_prompt.append(subprompt)
+        else:
+          converted_prompt.append(f"({subprompt}){weight}")
+    converted_prompt = " ".join(converted_prompt)
+    return converted_prompt
+
 def compel_encode_prompt(pipeline: typing.Any, *args, **kwargs):
    compel_encode_fn = COMPEL_ENCODE_FN_DICT.get(type(pipeline), None)
    if compel_encode_fn is None:
@@ -11,6 +27,12 @@ def compel_encode_prompt(pipeline: typing.Any, *args, **kwargs):
    return compel_encode_fn(pipeline, *args, **kwargs)
 
 def compel_encode_prompt_sdxl(pipeline: diffusers.StableDiffusionXLPipeline, prompt: str, negative_prompt: str, prompt_2: typing.Optional[str]=None, negative_prompt_2: typing.Optional[str]=None, refiner=False):
+    if shared.opts.data['prompt_attention'] != 'Compel parser':
+        prompt = convert_to_compel(prompt)
+        negative_prompt = convert_to_compel(negative_prompt)
+        prompt_2 = convert_to_compel(prompt_2)
+        negative_prompt_2 = convert_to_compel(negative_prompt_2)
+
     compel_te1 = Compel(
       tokenizer=pipeline.tokenizer,
       text_encoder=pipeline.text_encoder,
@@ -24,7 +46,7 @@ def compel_encode_prompt_sdxl(pipeline: diffusers.StableDiffusionXLPipeline, pro
       returned_embeddings_type=ReturnedEmbeddingsType.PENULTIMATE_HIDDEN_STATES_NON_NORMALIZED,
       requires_pooled=True,
       )
-    if refiner is None:
+    if refiner is False:
       positive_te1 = compel_te1(prompt)
       positive_te2, pooled = compel_te2(prompt_2)
       positive = torch.cat((positive_te1, positive_te2), dim=-1)
@@ -36,7 +58,7 @@ def compel_encode_prompt_sdxl(pipeline: diffusers.StableDiffusionXLPipeline, pro
       positive, pooled = compel_te2(prompt)
       negative, negative_pooled = compel_te2(negative_prompt)
 
-
+    shared.log.debug(compel_te1.parse_prompt_string(prompt))
     [prompt_embed, negative_embed] = compel_te2.pad_conditioning_tensors_to_same_length([positive, negative])
     return prompt_embed, pooled, negative_embed, negative_pooled