add --sequential_guidance option for low-RAM tradeoff

ldm/invoke/CLI.py

@@ -60,6 +60,7 @@ def main():
     Globals.always_use_cpu = args.always_use_cpu
     Globals.internet_available = args.internet_available and check_internet()
     Globals.disable_xformers = not args.xformers
+    Globals.sequential_guidance = args.sequential_guidance
     Globals.ckpt_convert = args.ckpt_convert
 
     print(f">> Internet connectivity is {Globals.internet_available}")
@@ -749,7 +750,7 @@ def import_ckpt_model(
     base_name = Path(url_attachment_name(path_or_url)).name if is_a_url else Path(path_or_url).name
     default_name = Path(base_name).stem
     default_description = f"Imported model {default_name}"
-    
+
     model_name, model_description = _get_model_name_and_desc(
         manager,
         completer,
@@ -834,7 +835,7 @@ def _ask_for_config_file(model_path: Union[str,Path], completer, plural: bool=Fa
         '2': 'v2-inference-v.yaml',
         '3': 'v1-inpainting-inference.yaml',
     }
-    
+
     prompt = '''What type of models are these?:
 [1] Models based on Stable Diffusion 1.X
 [2] Models based on Stable Diffusion 2.X
@@ -843,7 +844,7 @@ def _ask_for_config_file(model_path: Union[str,Path], completer, plural: bool=Fa
 [1] A model based on Stable Diffusion 1.X
 [2] A model based on Stable Diffusion 2.X
 [3] An inpainting models based on Stable Diffusion 1.X
-[4] Something else''' 
+[4] Something else'''
     print(prompt)
     choice = input(f'Your choice: [{default}] ')
     choice = choice.strip() or default

ldm/invoke/args.py

@@ -91,14 +91,14 @@ import pydoc
 import re
 import shlex
 import sys
-import ldm.invoke
-import ldm.invoke.pngwriter
-
-from ldm.invoke.globals import Globals
-from ldm.invoke.prompt_parser import split_weighted_subprompts
 from argparse import Namespace
 from pathlib import Path
 
+import ldm.invoke
+import ldm.invoke.pngwriter
+from ldm.invoke.globals import Globals
+from ldm.invoke.prompt_parser import split_weighted_subprompts
+
 APP_ID = ldm.invoke.__app_id__
 APP_NAME = ldm.invoke.__app_name__
 APP_VERSION = ldm.invoke.__version__
@@ -484,6 +484,13 @@ class Args(object):
             action='store_true',
             help='Force free gpu memory before final decoding',
         )
+        model_group.add_argument(
+            '--sequential_guidance',
+            dest='sequential_guidance',
+            action='store_true',
+            help="Calculate guidance in serial instead of in parallel, lowering memory requirement "
+                 "at the expense of speed",
+        )
         model_group.add_argument(
             '--xformers',
             action=argparse.BooleanOptionalAction,

ldm/invoke/globals.py

@@ -13,8 +13,8 @@ the attributes:
 
 import os
 import os.path as osp
-from pathlib import Path
 from argparse import Namespace
+from pathlib import Path
 from typing import Union
 
 Globals = Namespace()
@@ -48,6 +48,9 @@ Globals.internet_available = True
 # Whether to disable xformers
 Globals.disable_xformers = False
 
+# Low-memory tradeoff for guidance calculations.
+Globals.sequential_guidance = False
+
 # whether we are forcing full precision
 Globals.full_precision = False
 

ldm/models/diffusion/shared_invokeai_diffusion.py

@@ -8,6 +8,7 @@ import torch
 from diffusers.models.cross_attention import AttnProcessor
 from typing_extensions import TypeAlias
 
+from ldm.invoke.globals import Globals
 from ldm.models.diffusion.cross_attention_control import Arguments, \
     restore_default_cross_attention, override_cross_attention, Context, get_cross_attention_modules, \
     CrossAttentionType, SwapCrossAttnContext
@@ -35,7 +36,7 @@ class InvokeAIDiffuserComponent:
     * Hybrid conditioning (used for inpainting)
     '''
     debug_thresholding = False
-    sequential_conditioning = False
+    sequential_guidance = False
 
     @dataclass
     class ExtraConditioningInfo:
@@ -60,6 +61,7 @@ class InvokeAIDiffuserComponent:
         self.is_running_diffusers = is_running_diffusers
         self.model_forward_callback = model_forward_callback
         self.cross_attention_control_context = None
+        self.sequential_guidance = Globals.sequential_guidance
 
     @contextmanager
     def custom_attention_context(self,
@@ -154,7 +156,7 @@ class InvokeAIDiffuserComponent:
                                                                                                            unconditioning,
                                                                                                            conditioning,
                                                                                                            cross_attention_control_types_to_do)
-        elif self.sequential_conditioning:
+        elif self.sequential_guidance:
             unconditioned_next_x, conditioned_next_x = self._apply_standard_conditioning_sequentially(
                 x, sigma, unconditioning, conditioning)