InvokeAI/invokeai/app/invocations/compel.py

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

534 lines
20 KiB
Python
Raw Normal View History

2023-05-30 23:12:27 +00:00
import re
from dataclasses import dataclass
from typing import List, Union
import torch
2023-07-17 22:49:45 +00:00
from compel import Compel, ReturnedEmbeddingsType
from compel.prompt_parser import Blend, Conjunction, CrossAttentionControlSubstitute, FlattenedPrompt, Fragment
from invokeai.app.invocations.primitives import ConditioningField, ConditioningOutput
from invokeai.backend.stable_diffusion.diffusion.conditioning_data import (
BasicConditioningInfo,
ExtraConditioningInfo,
SDXLConditioningInfo,
)
2023-05-30 23:12:27 +00:00
from ...backend.model_management.lora import ModelPatcher
from ...backend.model_management.models import ModelNotFoundException, ModelType
from ...backend.util.devices import torch_dtype
from .baseinvocation import (
BaseInvocation,
BaseInvocationOutput,
FieldDescriptions,
Input,
InputField,
InvocationContext,
OutputField,
UIComponent,
invocation,
invocation_output,
)
from .model import ClipField
2023-07-27 14:54:01 +00:00
@dataclass
class ConditioningFieldData:
2023-08-08 20:33:52 +00:00
conditionings: List[BasicConditioningInfo]
# unconditioned: Optional[torch.Tensor]
2023-07-27 14:54:01 +00:00
# class ConditioningAlgo(str, Enum):
# Compose = "compose"
# ComposeEx = "compose_ex"
# PerpNeg = "perp_neg"
2023-07-27 14:54:01 +00:00
@invocation("compel", title="Prompt", tags=["prompt", "compel"], category="conditioning", version="1.0.0")
class CompelInvocation(BaseInvocation):
2023-05-05 18:09:29 +00:00
"""Parse prompt using compel package to conditioning."""
prompt: str = InputField(
default="",
description=FieldDescriptions.compel_prompt,
ui_component=UIComponent.Textarea,
)
clip: ClipField = InputField(
title="CLIP",
description=FieldDescriptions.clip,
input=Input.Connection,
)
@torch.no_grad()
def invoke(self, context: InvocationContext) -> ConditioningOutput:
tokenizer_info = context.services.model_manager.get_model(
**self.clip.tokenizer.dict(),
context=context,
)
2023-05-30 23:12:27 +00:00
text_encoder_info = context.services.model_manager.get_model(
**self.clip.text_encoder.dict(),
context=context,
2023-05-30 23:12:27 +00:00
)
def _lora_loader():
for lora in self.clip.loras:
lora_info = context.services.model_manager.get_model(**lora.dict(exclude={"weight"}), context=context)
yield (lora_info.context.model, lora.weight)
del lora_info
return
# loras = [(context.services.model_manager.get_model(**lora.dict(exclude={"weight"})).context.model, lora.weight) for lora in self.clip.loras]
ti_list = []
for trigger in re.findall(r"<[a-zA-Z0-9., _-]+>", self.prompt):
name = trigger[1:-1]
try:
ti_list.append(
2023-08-03 23:01:05 +00:00
(
name,
context.services.model_manager.get_model(
model_name=name,
base_model=self.clip.text_encoder.base_model,
model_type=ModelType.TextualInversion,
context=context,
).context.model,
)
2023-05-30 23:12:27 +00:00
)
2023-07-05 16:46:00 +00:00
except ModelNotFoundException:
# print(e)
# import traceback
2023-07-05 16:46:00 +00:00
# print(traceback.format_exc())
print(f'Warn: trigger: "{trigger}" not found')
with (
ModelPatcher.apply_lora_text_encoder(text_encoder_info.context.model, _lora_loader()),
ModelPatcher.apply_ti(tokenizer_info.context.model, text_encoder_info.context.model, ti_list) as (
tokenizer,
ti_manager,
),
ModelPatcher.apply_clip_skip(text_encoder_info.context.model, self.clip.skipped_layers),
text_encoder_info as text_encoder,
):
compel = Compel(
tokenizer=tokenizer,
text_encoder=text_encoder,
textual_inversion_manager=ti_manager,
dtype_for_device_getter=torch_dtype,
2023-07-30 12:20:59 +00:00
truncate_long_prompts=False,
)
2023-05-12 01:24:29 +00:00
conjunction = Compel.parse_prompt_string(self.prompt)
2023-05-12 01:24:29 +00:00
if context.services.configuration.log_tokenization:
2023-08-29 14:07:33 +00:00
log_tokenization_for_conjunction(conjunction, tokenizer)
2023-07-30 12:20:59 +00:00
c, options = compel.build_conditioning_tensor_for_conjunction(conjunction)
ec = ExtraConditioningInfo(
tokens_count_including_eos_bos=get_max_token_count(tokenizer, conjunction),
cross_attention_control_args=options.get("cross_attention_control", None),
)
c = c.detach().to("cpu")
2023-07-16 03:24:24 +00:00
conditioning_data = ConditioningFieldData(
conditionings=[
BasicConditioningInfo(
embeds=c,
extra_conditioning=ec,
)
]
)
conditioning_name = f"{context.graph_execution_state_id}_{self.id}_conditioning"
2023-07-16 03:24:24 +00:00
context.services.latents.save(conditioning_name, conditioning_data)
return ConditioningOutput(
conditioning=ConditioningField(
conditioning_name=conditioning_name,
),
)
2023-07-27 14:54:01 +00:00
2023-07-17 22:49:45 +00:00
class SDXLPromptInvocationBase:
def run_clip_compel(
self,
context: InvocationContext,
clip_field: ClipField,
prompt: str,
get_pooled: bool,
lora_prefix: str,
zero_on_empty: bool,
):
2023-07-17 22:49:45 +00:00
tokenizer_info = context.services.model_manager.get_model(
**clip_field.tokenizer.dict(),
context=context,
2023-07-17 22:49:45 +00:00
)
text_encoder_info = context.services.model_manager.get_model(
**clip_field.text_encoder.dict(),
context=context,
2023-07-17 22:49:45 +00:00
)
# return zero on empty
if prompt == "" and zero_on_empty:
cpu_text_encoder = text_encoder_info.context.model
c = torch.zeros(
(1, cpu_text_encoder.config.max_position_embeddings, cpu_text_encoder.config.hidden_size),
dtype=text_encoder_info.context.cache.precision,
2023-07-17 22:49:45 +00:00
)
if get_pooled:
c_pooled = torch.zeros(
(1, cpu_text_encoder.config.hidden_size),
dtype=c.dtype,
)
2023-07-17 22:49:45 +00:00
else:
c_pooled = None
return c, c_pooled, None
def _lora_loader():
for lora in clip_field.loras:
lora_info = context.services.model_manager.get_model(**lora.dict(exclude={"weight"}), context=context)
yield (lora_info.context.model, lora.weight)
del lora_info
return
# loras = [(context.services.model_manager.get_model(**lora.dict(exclude={"weight"})).context.model, lora.weight) for lora in self.clip.loras]
ti_list = []
2023-07-17 22:49:45 +00:00
for trigger in re.findall(r"<[a-zA-Z0-9., _-]+>", prompt):
name = trigger[1:-1]
try:
ti_list.append(
2023-08-03 23:01:05 +00:00
(
name,
context.services.model_manager.get_model(
model_name=name,
base_model=clip_field.text_encoder.base_model,
model_type=ModelType.TextualInversion,
context=context,
).context.model,
)
)
except ModelNotFoundException:
# print(e)
# import traceback
# print(traceback.format_exc())
print(f'Warn: trigger: "{trigger}" not found')
with (
ModelPatcher.apply_lora(text_encoder_info.context.model, _lora_loader(), lora_prefix),
ModelPatcher.apply_ti(tokenizer_info.context.model, text_encoder_info.context.model, ti_list) as (
tokenizer,
ti_manager,
),
ModelPatcher.apply_clip_skip(text_encoder_info.context.model, clip_field.skipped_layers),
text_encoder_info as text_encoder,
):
compel = Compel(
tokenizer=tokenizer,
text_encoder=text_encoder,
textual_inversion_manager=ti_manager,
dtype_for_device_getter=torch_dtype,
2023-07-30 12:20:59 +00:00
truncate_long_prompts=False, # TODO:
2023-07-17 22:49:45 +00:00
returned_embeddings_type=ReturnedEmbeddingsType.PENULTIMATE_HIDDEN_STATES_NON_NORMALIZED, # TODO: clip skip
requires_pooled=get_pooled,
)
2023-07-17 22:49:45 +00:00
conjunction = Compel.parse_prompt_string(prompt)
if context.services.configuration.log_tokenization:
2023-07-17 22:49:45 +00:00
# TODO: better logging for and syntax
2023-07-30 12:20:59 +00:00
log_tokenization_for_conjunction(conjunction, tokenizer)
2023-07-17 22:49:45 +00:00
# TODO: ask for optimizations? to not run text_encoder twice
c, options = compel.build_conditioning_tensor_for_conjunction(conjunction)
if get_pooled:
c_pooled = compel.conditioning_provider.get_pooled_embeddings([prompt])
else:
c_pooled = None
ec = ExtraConditioningInfo(
tokens_count_including_eos_bos=get_max_token_count(tokenizer, conjunction),
cross_attention_control_args=options.get("cross_attention_control", None),
)
del tokenizer
del text_encoder
del tokenizer_info
del text_encoder_info
c = c.detach().to("cpu")
if c_pooled is not None:
c_pooled = c_pooled.detach().to("cpu")
2023-07-17 22:49:45 +00:00
return c, c_pooled, ec
2023-07-27 14:54:01 +00:00
@invocation(
"sdxl_compel_prompt",
title="SDXL Prompt",
tags=["sdxl", "compel", "prompt"],
category="conditioning",
version="1.0.0",
)
2023-07-17 22:49:45 +00:00
class SDXLCompelPromptInvocation(BaseInvocation, SDXLPromptInvocationBase):
"""Parse prompt using compel package to conditioning."""
prompt: str = InputField(default="", description=FieldDescriptions.compel_prompt, ui_component=UIComponent.Textarea)
style: str = InputField(default="", description=FieldDescriptions.compel_prompt, ui_component=UIComponent.Textarea)
original_width: int = InputField(default=1024, description="")
original_height: int = InputField(default=1024, description="")
crop_top: int = InputField(default=0, description="")
crop_left: int = InputField(default=0, description="")
target_width: int = InputField(default=1024, description="")
target_height: int = InputField(default=1024, description="")
clip: ClipField = InputField(description=FieldDescriptions.clip, input=Input.Connection, title="CLIP 1")
clip2: ClipField = InputField(description=FieldDescriptions.clip, input=Input.Connection, title="CLIP 2")
@torch.no_grad()
def invoke(self, context: InvocationContext) -> ConditioningOutput:
2023-08-13 09:28:39 +00:00
c1, c1_pooled, ec1 = self.run_clip_compel(
2023-08-13 16:31:14 +00:00
context, self.clip, self.prompt, False, "lora_te1_", zero_on_empty=True
2023-08-13 09:28:39 +00:00
)
2023-07-17 22:49:45 +00:00
if self.style.strip() == "":
2023-08-13 09:28:39 +00:00
c2, c2_pooled, ec2 = self.run_clip_compel(
context, self.clip2, self.prompt, True, "lora_te2_", zero_on_empty=True
)
2023-07-17 22:49:45 +00:00
else:
2023-08-13 09:28:39 +00:00
c2, c2_pooled, ec2 = self.run_clip_compel(
context, self.clip2, self.style, True, "lora_te2_", zero_on_empty=True
)
2023-07-17 22:49:45 +00:00
original_size = (self.original_height, self.original_width)
crop_coords = (self.crop_top, self.crop_left)
target_size = (self.target_height, self.target_width)
add_time_ids = torch.tensor([original_size + crop_coords + target_size])
# [1, 77, 768], [1, 154, 1280]
if c1.shape[1] < c2.shape[1]:
c1 = torch.cat(
[
c1,
torch.zeros(
(c1.shape[0], c2.shape[1] - c1.shape[1], c1.shape[2]), device=c1.device, dtype=c1.dtype
),
],
dim=1,
)
elif c1.shape[1] > c2.shape[1]:
c2 = torch.cat(
[
c2,
torch.zeros(
(c2.shape[0], c1.shape[1] - c2.shape[1], c2.shape[2]), device=c2.device, dtype=c2.dtype
),
],
dim=1,
)
conditioning_data = ConditioningFieldData(
conditionings=[
SDXLConditioningInfo(
embeds=torch.cat([c1, c2], dim=-1),
pooled_embeds=c2_pooled,
2023-07-17 22:49:45 +00:00
add_time_ids=add_time_ids,
extra_conditioning=ec1,
)
]
)
conditioning_name = f"{context.graph_execution_state_id}_{self.id}_conditioning"
context.services.latents.save(conditioning_name, conditioning_data)
return ConditioningOutput(
conditioning=ConditioningField(
conditioning_name=conditioning_name,
),
)
2023-07-27 14:54:01 +00:00
@invocation(
"sdxl_refiner_compel_prompt",
title="SDXL Refiner Prompt",
tags=["sdxl", "compel", "prompt"],
category="conditioning",
version="1.0.0",
)
2023-07-17 22:49:45 +00:00
class SDXLRefinerCompelPromptInvocation(BaseInvocation, SDXLPromptInvocationBase):
"""Parse prompt using compel package to conditioning."""
style: str = InputField(
default="", description=FieldDescriptions.compel_prompt, ui_component=UIComponent.Textarea
) # TODO: ?
original_width: int = InputField(default=1024, description="")
original_height: int = InputField(default=1024, description="")
crop_top: int = InputField(default=0, description="")
crop_left: int = InputField(default=0, description="")
aesthetic_score: float = InputField(default=6.0, description=FieldDescriptions.sdxl_aesthetic)
clip2: ClipField = InputField(description=FieldDescriptions.clip, input=Input.Connection)
@torch.no_grad()
def invoke(self, context: InvocationContext) -> ConditioningOutput:
2023-07-31 20:18:02 +00:00
# TODO: if there will appear lora for refiner - write proper prefix
c2, c2_pooled, ec2 = self.run_clip_compel(context, self.clip2, self.style, True, "<NONE>", zero_on_empty=False)
original_size = (self.original_height, self.original_width)
crop_coords = (self.crop_top, self.crop_left)
add_time_ids = torch.tensor([original_size + crop_coords + (self.aesthetic_score,)])
conditioning_data = ConditioningFieldData(
conditionings=[
SDXLConditioningInfo(
embeds=c2,
pooled_embeds=c2_pooled,
add_time_ids=add_time_ids,
extra_conditioning=ec2, # or None
)
]
)
conditioning_name = f"{context.graph_execution_state_id}_{self.id}_conditioning"
context.services.latents.save(conditioning_name, conditioning_data)
return ConditioningOutput(
conditioning=ConditioningField(
conditioning_name=conditioning_name,
),
)
2023-07-17 22:49:45 +00:00
@invocation_output("clip_skip_output")
2023-07-06 14:39:49 +00:00
class ClipSkipInvocationOutput(BaseInvocationOutput):
"""Clip skip node output"""
2023-07-27 14:54:01 +00:00
clip: ClipField = OutputField(default=None, description=FieldDescriptions.clip, title="CLIP")
2023-07-06 14:39:49 +00:00
2023-07-27 14:54:01 +00:00
@invocation("clip_skip", title="CLIP Skip", tags=["clipskip", "clip", "skip"], category="conditioning", version="1.0.0")
2023-07-06 14:39:49 +00:00
class ClipSkipInvocation(BaseInvocation):
"""Skip layers in clip text_encoder model."""
2023-07-27 14:54:01 +00:00
clip: ClipField = InputField(description=FieldDescriptions.clip, input=Input.Connection, title="CLIP")
skipped_layers: int = InputField(default=0, description=FieldDescriptions.skipped_layers)
2023-07-18 14:26:45 +00:00
2023-07-06 14:39:49 +00:00
def invoke(self, context: InvocationContext) -> ClipSkipInvocationOutput:
self.clip.skipped_layers += self.skipped_layers
return ClipSkipInvocationOutput(
clip=self.clip,
)
def get_max_token_count(
tokenizer, prompt: Union[FlattenedPrompt, Blend, Conjunction], truncate_if_too_long=False
) -> int:
if type(prompt) is Blend:
blend: Blend = prompt
return max([get_max_token_count(tokenizer, p, truncate_if_too_long) for p in blend.prompts])
elif type(prompt) is Conjunction:
conjunction: Conjunction = prompt
return sum([get_max_token_count(tokenizer, p, truncate_if_too_long) for p in conjunction.prompts])
else:
return len(get_tokens_for_prompt_object(tokenizer, prompt, truncate_if_too_long))
def get_tokens_for_prompt_object(tokenizer, parsed_prompt: FlattenedPrompt, truncate_if_too_long=True) -> List[str]:
if type(parsed_prompt) is Blend:
raise ValueError("Blend is not supported here - you need to get tokens for each of its .children")
text_fragments = [
(
x.text
if type(x) is Fragment
else (" ".join([f.text for f in x.original]) if type(x) is CrossAttentionControlSubstitute else str(x))
)
for x in parsed_prompt.children
]
text = " ".join(text_fragments)
tokens = tokenizer.tokenize(text)
if truncate_if_too_long:
max_tokens_length = tokenizer.model_max_length - 2 # typically 75
tokens = tokens[0:max_tokens_length]
return tokens
def log_tokenization_for_conjunction(c: Conjunction, tokenizer, display_label_prefix=None):
display_label_prefix = display_label_prefix or ""
for i, p in enumerate(c.prompts):
if len(c.prompts) > 1:
this_display_label_prefix = f"{display_label_prefix}(conjunction part {i + 1}, weight={c.weights[i]})"
else:
this_display_label_prefix = display_label_prefix
log_tokenization_for_prompt_object(p, tokenizer, display_label_prefix=this_display_label_prefix)
def log_tokenization_for_prompt_object(p: Union[Blend, FlattenedPrompt], tokenizer, display_label_prefix=None):
display_label_prefix = display_label_prefix or ""
if type(p) is Blend:
blend: Blend = p
for i, c in enumerate(blend.prompts):
log_tokenization_for_prompt_object(
c,
tokenizer,
display_label_prefix=f"{display_label_prefix}(blend part {i + 1}, weight={blend.weights[i]})",
)
elif type(p) is FlattenedPrompt:
flattened_prompt: FlattenedPrompt = p
if flattened_prompt.wants_cross_attention_control:
original_fragments = []
edited_fragments = []
for f in flattened_prompt.children:
if type(f) is CrossAttentionControlSubstitute:
original_fragments += f.original
edited_fragments += f.edited
else:
original_fragments.append(f)
edited_fragments.append(f)
original_text = " ".join([x.text for x in original_fragments])
log_tokenization_for_text(
original_text,
tokenizer,
display_label=f"{display_label_prefix}(.swap originals)",
)
edited_text = " ".join([x.text for x in edited_fragments])
log_tokenization_for_text(
edited_text,
tokenizer,
display_label=f"{display_label_prefix}(.swap replacements)",
)
else:
text = " ".join([x.text for x in flattened_prompt.children])
log_tokenization_for_text(text, tokenizer, display_label=display_label_prefix)
def log_tokenization_for_text(text, tokenizer, display_label=None, truncate_if_too_long=False):
"""shows how the prompt is tokenized
# usually tokens have '</w>' to indicate end-of-word,
# but for readability it has been replaced with ' '
"""
tokens = tokenizer.tokenize(text)
tokenized = ""
discarded = ""
usedTokens = 0
totalTokens = len(tokens)
for i in range(0, totalTokens):
token = tokens[i].replace("</w>", " ")
# alternate color
s = (usedTokens % 6) + 1
if truncate_if_too_long and i >= tokenizer.model_max_length:
discarded = discarded + f"\x1b[0;3{s};40m{token}"
else:
tokenized = tokenized + f"\x1b[0;3{s};40m{token}"
usedTokens += 1
if usedTokens > 0:
print(f'\n>> [TOKENLOG] Tokens {display_label or ""} ({usedTokens}):')
print(f"{tokenized}\x1b[0m")
if discarded != "":
print(f"\n>> [TOKENLOG] Tokens Discarded ({totalTokens - usedTokens}):")
print(f"{discarded}\x1b[0m")