Artificial Sweetener
017e3fc7fe
Keep compatible parallel regional paths on one inherited model lineage, including NegPip interoperability, while retaining bounded fused and optional Triton execution paths. Expand graph-shape, lifecycle, memory-safety, and runtime regressions across the supported attention families.
286 lines
11 KiB
Python
286 lines
11 KiB
Python
# SimpleSyrup - workflow-focused ComfyUI extensions for image generation
|
|
# Copyright (C) 2026 Artificial Sweetener and contributors
|
|
# SPDX-License-Identifier: AGPL-3.0-or-later
|
|
|
|
"""Process regional Attention Coupling contexts through installed Comfy APIs."""
|
|
|
|
from __future__ import annotations
|
|
|
|
from collections.abc import Callable
|
|
from typing import cast
|
|
from uuid import UUID
|
|
|
|
import torch
|
|
from comfy import sampler_helpers, samplers
|
|
|
|
from ..domain.conditioning_schedule import ConditioningScheduleRange
|
|
from ..domain.processed_regional_attention import (
|
|
ProcessedRegionalAttentionBranch,
|
|
ProcessedRegionalAttentionContext,
|
|
ProcessedRegionalAttentionEntry,
|
|
ProcessedRegionalAttentionPlan,
|
|
)
|
|
from ..domain.raw_regional_attention import (
|
|
RawRegionalAttentionBranch,
|
|
)
|
|
from ..services.attention_coupling_preparation_service import (
|
|
AttentionCouplingPreparation,
|
|
)
|
|
from .attention_coupling.context_validation import RegionalContextValidator
|
|
from .ppm_negpip_interop import PpmNegpipInterop
|
|
|
|
|
|
class ComfyRegionalConditioningProcessor:
|
|
"""Convert every raw context through the same model path as Comfy sampling."""
|
|
|
|
def process(
|
|
self,
|
|
preparation: AttentionCouplingPreparation,
|
|
*,
|
|
model: object,
|
|
noise: torch.Tensor,
|
|
device: torch.device,
|
|
context_validator: RegionalContextValidator,
|
|
negpip: PpmNegpipInterop | None = None,
|
|
) -> ProcessedRegionalAttentionPlan:
|
|
"""Return model-ready positive and negative context banks."""
|
|
|
|
if not isinstance(preparation, AttentionCouplingPreparation):
|
|
raise TypeError("Regional context processing requires a preparation.")
|
|
if not isinstance(noise, torch.Tensor) or noise.ndim not in (4, 5):
|
|
raise TypeError(
|
|
"Regional context processing noise must be a 4D or 5D tensor."
|
|
)
|
|
if int(noise.shape[0]) < 1:
|
|
raise ValueError(
|
|
"Regional context processing noise batch must be positive."
|
|
)
|
|
if not isinstance(device, torch.device):
|
|
raise TypeError("Regional context processing device must be torch.device.")
|
|
if not isinstance(context_validator, RegionalContextValidator):
|
|
raise TypeError("Regional context validator has an invalid type.")
|
|
if negpip is not None and not isinstance(negpip, PpmNegpipInterop):
|
|
raise TypeError("Regional conditioning NegPiP state has an invalid type.")
|
|
base_model = getattr(model, "model", None)
|
|
extra_conds = getattr(base_model, "extra_conds", None)
|
|
if not callable(extra_conds):
|
|
raise TypeError(
|
|
"MODEL must expose model.extra_conds for context processing."
|
|
)
|
|
model_function = cast(Callable[..., dict[str, object]], extra_conds)
|
|
|
|
self._preflight_branch("positive", preparation.plan.positive)
|
|
self._preflight_branch("negative", preparation.plan.negative)
|
|
positive = self._process_branch(
|
|
preparation.plan.positive,
|
|
prompt_type="positive",
|
|
model=base_model,
|
|
model_function=model_function,
|
|
noise=noise,
|
|
device=device,
|
|
context_validator=context_validator,
|
|
negpip=negpip,
|
|
)
|
|
negative = self._process_branch(
|
|
preparation.plan.negative,
|
|
prompt_type="negative",
|
|
model=base_model,
|
|
model_function=model_function,
|
|
noise=noise,
|
|
device=device,
|
|
context_validator=context_validator,
|
|
negpip=negpip,
|
|
)
|
|
return ProcessedRegionalAttentionPlan(
|
|
positive=positive,
|
|
negative=negative,
|
|
mask_bank=preparation.plan.mask_bank,
|
|
lora_plan=preparation.plan.lora_plan,
|
|
)
|
|
|
|
def _process_branch(
|
|
self,
|
|
branch: RawRegionalAttentionBranch,
|
|
*,
|
|
prompt_type: str,
|
|
model: object,
|
|
model_function: Callable[..., dict[str, object]],
|
|
noise: torch.Tensor,
|
|
device: torch.device,
|
|
context_validator: RegionalContextValidator,
|
|
negpip: PpmNegpipInterop | None,
|
|
) -> ProcessedRegionalAttentionBranch:
|
|
"""Process one base plus its ordered regional context bank."""
|
|
|
|
base = self._process_context(
|
|
branch.base_conditioning,
|
|
conditioning_index=0,
|
|
region_index=None,
|
|
prompt_type=prompt_type,
|
|
model=model,
|
|
model_function=model_function,
|
|
noise=noise,
|
|
device=device,
|
|
context_validator=context_validator,
|
|
negpip=negpip,
|
|
)
|
|
regional = tuple(
|
|
self._process_context(
|
|
context.conditioning,
|
|
conditioning_index=context.conditioning_index,
|
|
region_index=context.region_index,
|
|
prompt_type=prompt_type,
|
|
model=model,
|
|
model_function=model_function,
|
|
noise=noise,
|
|
device=device,
|
|
context_validator=context_validator,
|
|
negpip=negpip,
|
|
)
|
|
for context in branch.regional_contexts
|
|
)
|
|
return ProcessedRegionalAttentionBranch(base, regional)
|
|
|
|
def _process_context(
|
|
self,
|
|
conditioning: object,
|
|
*,
|
|
conditioning_index: int,
|
|
region_index: int | None,
|
|
prompt_type: str,
|
|
model: object,
|
|
model_function: Callable[..., dict[str, object]],
|
|
noise: torch.Tensor,
|
|
device: torch.device,
|
|
context_validator: RegionalContextValidator,
|
|
negpip: PpmNegpipInterop | None,
|
|
) -> ProcessedRegionalAttentionContext:
|
|
"""Convert and extract one exact post-adapter Anima context tensor."""
|
|
|
|
converted = sampler_helpers.convert_cond(conditioning)
|
|
samplers.calculate_start_end_timesteps(model, converted)
|
|
encoded = samplers.encode_model_conds(
|
|
model_function,
|
|
converted,
|
|
noise,
|
|
device,
|
|
prompt_type,
|
|
)
|
|
if not isinstance(encoded, list) or not encoded:
|
|
raise ValueError(
|
|
f"{prompt_type} conditioning {conditioning_index} must convert "
|
|
"to at least one Comfy condition."
|
|
)
|
|
entries = tuple(
|
|
self._process_entry(
|
|
encoded_item,
|
|
entry_index=entry_index,
|
|
conditioning_index=conditioning_index,
|
|
prompt_type=prompt_type,
|
|
context_validator=context_validator,
|
|
negpip=negpip,
|
|
)
|
|
for entry_index, encoded_item in enumerate(encoded)
|
|
)
|
|
return ProcessedRegionalAttentionContext(
|
|
conditioning_index=conditioning_index,
|
|
region_index=region_index,
|
|
entries=entries,
|
|
)
|
|
|
|
@staticmethod
|
|
def _process_entry(
|
|
encoded_item: object,
|
|
*,
|
|
entry_index: int,
|
|
conditioning_index: int,
|
|
prompt_type: str,
|
|
context_validator: RegionalContextValidator,
|
|
negpip: PpmNegpipInterop | None,
|
|
) -> ProcessedRegionalAttentionEntry:
|
|
"""Extract one exact post-adapter Anima context and Comfy strength."""
|
|
|
|
if not isinstance(encoded_item, dict):
|
|
raise TypeError("Comfy encoded conditioning item must be a dictionary.")
|
|
model_conds = encoded_item.get("model_conds")
|
|
if not isinstance(model_conds, dict):
|
|
raise TypeError("Comfy encoded conditioning must contain model_conds.")
|
|
cross_attention = model_conds.get("c_crossattn")
|
|
context = getattr(cross_attention, "cond", None)
|
|
if not isinstance(context, torch.Tensor):
|
|
raise TypeError(
|
|
f"{prompt_type} conditioning {conditioning_index} did not produce "
|
|
"a tensor c_crossattn condition."
|
|
)
|
|
if context.ndim != 3:
|
|
raise ValueError(
|
|
f"{prompt_type} conditioning {conditioning_index} c_crossattn "
|
|
"must use BxSxD layout."
|
|
)
|
|
context_validator.validate(
|
|
context,
|
|
prompt_type=prompt_type,
|
|
conditioning_index=conditioning_index,
|
|
)
|
|
strength = encoded_item.get("strength", 1.0)
|
|
if isinstance(strength, bool) or not isinstance(strength, int | float):
|
|
raise TypeError(
|
|
f"{prompt_type} conditioning {conditioning_index} entry "
|
|
f"{entry_index} strength must be a real number."
|
|
)
|
|
entry_uuid = encoded_item.get("uuid")
|
|
if not isinstance(entry_uuid, UUID):
|
|
raise TypeError(
|
|
f"{prompt_type} conditioning {conditioning_index} entry "
|
|
f"{entry_index} must retain a Comfy UUID."
|
|
)
|
|
return ProcessedRegionalAttentionEntry(
|
|
entry_index=entry_index,
|
|
uuid=entry_uuid,
|
|
schedule=ConditioningScheduleRange(
|
|
_optional_boundary(encoded_item, "start_percent"),
|
|
_optional_boundary(encoded_item, "end_percent"),
|
|
_optional_boundary(encoded_item, "timestep_start"),
|
|
_optional_boundary(encoded_item, "timestep_end"),
|
|
),
|
|
cross_attention=context,
|
|
strength=float(strength),
|
|
cross_attention_value_multiplier=(
|
|
None
|
|
if negpip is None
|
|
else negpip.extract_value_multiplier(model_conds, context)
|
|
),
|
|
)
|
|
|
|
@staticmethod
|
|
def _preflight_branch(
|
|
prompt_type: str,
|
|
branch: RawRegionalAttentionBranch,
|
|
) -> None:
|
|
"""Require every authored context to contain at least one Comfy entry."""
|
|
|
|
indexed = ((0, branch.base_conditioning),) + tuple(
|
|
(context.conditioning_index, context.conditioning)
|
|
for context in branch.regional_contexts
|
|
)
|
|
for conditioning_index, conditioning in indexed:
|
|
if not isinstance(conditioning, list) or not conditioning:
|
|
raise ValueError(
|
|
f"{prompt_type} conditioning {conditioning_index} must contain "
|
|
"at least one Comfy condition."
|
|
)
|
|
|
|
|
|
COMFY_REGIONAL_CONDITIONING_PROCESSOR = ComfyRegionalConditioningProcessor()
|
|
|
|
|
|
def _optional_boundary(values: dict[str, object], key: str) -> float | None:
|
|
"""Narrow one optional Comfy schedule boundary for the domain owner."""
|
|
|
|
value = values.get(key)
|
|
if value is None:
|
|
return None
|
|
if isinstance(value, bool) or not isinstance(value, int | float):
|
|
raise TypeError(f"Comfy conditioning {key} must be a real number.")
|
|
return float(value)
|