Merge ac8c05398bd103404995fafb936e88a1b711344f into 374bb6cc384d2a19422c0b07d69de0a41d1f3f4d

modules/processing.py

@@ -16,7 +16,7 @@ from skimage import exposure
 from typing import Any
 
 import modules.sd_hijack
-from modules import devices, prompt_parser, masking, sd_samplers, lowvram, infotext_utils, extra_networks, sd_vae_approx, scripts, sd_samplers_common, sd_unet, errors, rng, profiling
+from modules import devices, prompt_parser, masking, sd_samplers, lowvram, infotext_utils, extra_networks, sd_vae_approx, scripts, sd_samplers_common, sd_unet, errors, rng, profiling, util
 from modules.rng import slerp # noqa: F401
 from modules.sd_hijack import model_hijack
 from modules.sd_samplers_common import images_tensor_to_samples, decode_first_stage, approximation_indexes
@@ -457,6 +457,20 @@ class StableDiffusionProcessing:
             opts.emphasis,
         )
 
+    def apply_generation_params_list(self, generation_params_states):
+        """add and apply generation_params_states to self.extra_generation_params"""
+        for key, value in generation_params_states.items():
+            if key in self.extra_generation_params and isinstance(current_value := self.extra_generation_params[key], util.GenerationParametersList):
+                self.extra_generation_params[key] = current_value + value
+            else:
+                self.extra_generation_params[key] = value
+
+    def clear_marked_generation_params(self):
+        """clears any generation parameters that are with the attribute to_be_clear_before_batch = True"""
+        for key, value in list(self.extra_generation_params.items()):
+            if getattr(value, 'to_be_clear_before_batch', False):
+                self.extra_generation_params.pop(key)
+
     def get_conds_with_caching(self, function, required_prompts, steps, caches, extra_network_data, hires_steps=None):
         """
         Returns the result of calling function(shared.sd_model, required_prompts, steps)
@@ -480,6 +494,10 @@ class StableDiffusionProcessing:
 
         for cache in caches:
             if cache[0] is not None and cached_params == cache[0]:
+                if len(cache) == 3:
+                    generation_params_states, cached_cached_params = cache[2]
+                    if cached_params == cached_cached_params:
+                        self.apply_generation_params_list(generation_params_states)
                 return cache[1]
 
         cache = caches[0]
@@ -487,6 +505,13 @@ class StableDiffusionProcessing:
         with devices.autocast():
             cache[1] = function(shared.sd_model, required_prompts, steps, hires_steps, shared.opts.use_old_scheduling)
 
+        generation_params_states = model_hijack.extract_generation_params_states()
+        self.apply_generation_params_list(generation_params_states)
+        if len(cache) == 2:
+            cache.append((generation_params_states, cached_params))
+        else:
+            cache[2] = (generation_params_states, cached_params)
+
         cache[0] = cached_params
         return cache[1]
 
@@ -502,6 +527,8 @@ class StableDiffusionProcessing:
         self.uc = self.get_conds_with_caching(prompt_parser.get_learned_conditioning, negative_prompts, total_steps, [self.cached_uc], self.extra_network_data)
         self.c = self.get_conds_with_caching(prompt_parser.get_multicond_learned_conditioning, prompts, total_steps, [self.cached_c], self.extra_network_data)
 
+        self.extra_generation_params.update(model_hijack.extra_generation_params)
+
     def get_conds(self):
         return self.c, self.uc
 
@@ -801,10 +828,10 @@ def create_infotext(p, all_prompts, all_seeds, all_subseeds, comments=None, iter
 
     for key, value in generation_params.items():
         try:
-            if isinstance(value, list):
-                generation_params[key] = value[index]
-            elif callable(value):
+            if callable(value):
                 generation_params[key] = value(**locals())
+            elif isinstance(value, list):
+                generation_params[key] = value[index]
         except Exception:
             errors.report(f'Error creating infotext for key "{key}"', exc_info=True)
             generation_params[key] = None
@@ -938,6 +965,7 @@ def process_images_inner(p: StableDiffusionProcessing) -> Processed:
             if state.interrupted or state.stopping_generation:
                 break
 
+            p.clear_marked_generation_params()  # clean up some generation params are tagged to be cleared before batch
             sd_models.reload_model_weights()  # model can be changed for example by refiner
 
             p.prompts = p.all_prompts[n * p.batch_size:(n + 1) * p.batch_size]
@@ -965,8 +993,6 @@ def process_images_inner(p: StableDiffusionProcessing) -> Processed:
 
             p.setup_conds()
 
-            p.extra_generation_params.update(model_hijack.extra_generation_params)
-
             # params.txt should be saved after scripts.process_batch, since the
             # infotext could be modified by that callback
             # Example: a wildcard processed by process_batch sets an extra model
@@ -1513,6 +1539,8 @@ class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing):
         self.hr_uc = self.get_conds_with_caching(prompt_parser.get_learned_conditioning, hr_negative_prompts, self.firstpass_steps, [self.cached_hr_uc, self.cached_uc], self.hr_extra_network_data, total_steps)
         self.hr_c = self.get_conds_with_caching(prompt_parser.get_multicond_learned_conditioning, hr_prompts, self.firstpass_steps, [self.cached_hr_c, self.cached_c], self.hr_extra_network_data, total_steps)
 
+        self.extra_generation_params.update(model_hijack.extra_generation_params)
+
     def setup_conds(self):
         if self.is_hr_pass:
             # if we are in hr pass right now, the call is being made from the refiner, and we don't need to setup firstpass cons or switch model

modules/sd_hijack.py

@@ -2,7 +2,7 @@ import torch
 from torch.nn.functional import silu
 from types import MethodType
 
-from modules import devices, sd_hijack_optimizations, shared, script_callbacks, errors, sd_unet, patches
+from modules import devices, sd_hijack_optimizations, shared, script_callbacks, errors, sd_unet, patches, util
 from modules.hypernetworks import hypernetwork
 from modules.shared import cmd_opts
 from modules import sd_hijack_clip, sd_hijack_open_clip, sd_hijack_unet, sd_hijack_xlmr, xlmr, xlmr_m18
@@ -321,6 +321,14 @@ class StableDiffusionModelHijack:
         self.comments = []
         self.extra_generation_params = {}
 
+    def extract_generation_params_states(self):
+        """Extracts GenerationParametersList so that they can be cached and restored later"""
+        states = {}
+        for key in list(self.extra_generation_params):
+            if isinstance(self.extra_generation_params[key], util.GenerationParametersList):
+                states[key] = self.extra_generation_params.pop(key)
+        return states
+
     def get_prompt_lengths(self, text):
         if self.clip is None:
             return "-", "-"

modules/sd_hijack_clip.py

@@ -3,7 +3,7 @@ from collections import namedtuple
 
 import torch
 
-from modules import prompt_parser, devices, sd_hijack, sd_emphasis
+from modules import prompt_parser, devices, sd_hijack, sd_emphasis, util
 from modules.shared import opts
 
 
@@ -27,6 +27,30 @@ chunk. Those objects are found in PromptChunk.fixes and, are placed into FrozenC
 are applied by sd_hijack.EmbeddingsWithFixes's forward function."""
 
 
+class EmphasisMode(util.GenerationParametersList):
+    def __init__(self, emphasis_mode:str = None):
+        super().__init__()
+        self.emphasis_mode = emphasis_mode
+
+    def __call__(self, *args, **kwargs):
+        return self.emphasis_mode
+
+    def __add__(self, other):
+        if isinstance(other, EmphasisMode):
+            return self if self.emphasis_mode else other
+        elif isinstance(other, str):
+            return self.__str__() + other
+        return NotImplemented
+
+    def __radd__(self, other):
+        if isinstance(other, str):
+            return other + self.__str__()
+        return NotImplemented
+
+    def __str__(self):
+        return self.emphasis_mode if self.emphasis_mode else ''
+
+
 class TextConditionalModel(torch.nn.Module):
     def __init__(self):
         super().__init__()
@@ -238,12 +262,10 @@ class TextConditionalModel(torch.nn.Module):
                 hashes.append(f"{name}: {shorthash}")
 
             if hashes:
-                if self.hijack.extra_generation_params.get("TI hashes"):
-                    hashes.append(self.hijack.extra_generation_params.get("TI hashes"))
-                self.hijack.extra_generation_params["TI hashes"] = ", ".join(hashes)
+                self.hijack.extra_generation_params["TI hashes"] = util.GenerationParametersList(hashes)
 
-        if any(x for x in texts if "(" in x or "[" in x) and opts.emphasis != "Original":
-            self.hijack.extra_generation_params["Emphasis"] = opts.emphasis
+        if opts.emphasis != 'Original' and any(x for x in texts if '(' in x or '[' in x):
+            self.hijack.extra_generation_params["Emphasis"] = EmphasisMode(opts.emphasis)
 
         if self.return_pooled:
             return torch.hstack(zs), zs[0].pooled

modules/util.py

@@ -289,3 +289,49 @@ def compare_sha256(file_path: str, hash_prefix: str) -> bool:
         for chunk in iter(lambda: f.read(blksize), b""):
             hash_sha256.update(chunk)
     return hash_sha256.hexdigest().startswith(hash_prefix.strip().lower())
+
+
+class GenerationParametersList(list):
+    """A special object used in sd_hijack.StableDiffusionModelHijack for setting extra_generation_params
+    due to StableDiffusionProcessing.get_conds_with_caching
+    extra_generation_params set in StableDiffusionModelHijack will be lost when cached is used
+
+    When an extra_generation_params is set in StableDiffusionModelHijack using this object,
+    the params will be extracted by StableDiffusionModelHijack.extract_generation_params_states
+    the extracted params will be cached in StableDiffusionProcessing.get_conds_with_caching
+    and applyed to StableDiffusionProcessing.extra_generation_params by StableDiffusionProcessing.apply_generation_params_states
+
+    Example see modules.sd_hijack_clip.TextConditionalModel.hijack.extra_generation_params 'TI hashes' 'Emphasis'
+
+    Depending on the use case the methods can be overwritten.
+    In general __call__ method should return str or None, as normally it's called in modules.processing.create_infotext.
+    When called by create_infotext it will access to the locals() of the caller,
+    if return str, the value will be written to infotext, if return None will be ignored.
+    """
+
+    def __init__(self, *args, to_be_clear_before_batch=True, **kwargs):
+        super().__init__(*args, **kwargs)
+        self._to_be_clear_before_batch = to_be_clear_before_batch
+
+    def __call__(self, *args, **kwargs):
+        return ', '.join(sorted(set(self), key=natural_sort_key))
+
+    @property
+    def to_be_clear_before_batch(self):
+        return self._to_be_clear_before_batch
+
+    def __add__(self, other):
+        if isinstance(other, GenerationParametersList):
+            return self.__class__([*self, *other])
+        elif isinstance(other, str):
+            return self.__str__() + other
+        return NotImplemented
+
+    def __radd__(self, other):
+        if isinstance(other, str):
+            return other + self.__str__()
+        return NotImplemented
+
+    def __str__(self):
+        return self.__call__()
+