Merge pull request #12371 from AUTOMATIC1111/refiner

initial refiner support

modules/processing.py

@@ -377,6 +377,9 @@ class StableDiffusionProcessing:
         self.uc = self.get_conds_with_caching(prompt_parser.get_learned_conditioning, negative_prompts, self.steps * self.step_multiplier, [self.cached_uc], self.extra_network_data)
         self.c = self.get_conds_with_caching(prompt_parser.get_multicond_learned_conditioning, prompts, self.steps * self.step_multiplier, [self.cached_c], self.extra_network_data)
 
+    def get_conds(self):
+        return self.c, self.uc
+
     def parse_extra_network_prompts(self):
         self.prompts, self.extra_network_data = extra_networks.parse_prompts(self.prompts)
 
@@ -611,6 +614,10 @@ def process_images(p: StableDiffusionProcessing) -> Processed:
     stored_opts = {k: opts.data[k] for k in p.override_settings.keys()}
 
     try:
+        # after running refiner, the refiner model is not unloaded - webui swaps back to main model here
+        if shared.sd_model.sd_checkpoint_info.title != opts.sd_model_checkpoint:
+            sd_models.reload_model_weights()
+
         # if no checkpoint override or the override checkpoint can't be found, remove override entry and load opts checkpoint
         if sd_models.checkpoint_aliases.get(p.override_settings.get('sd_model_checkpoint')) is None:
             p.override_settings.pop('sd_model_checkpoint', None)
@@ -710,6 +717,8 @@ def process_images_inner(p: StableDiffusionProcessing) -> Processed:
             if state.interrupted:
                 break
 
+            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]
             p.negative_prompts = p.all_negative_prompts[n * p.batch_size:(n + 1) * p.batch_size]
             p.seeds = p.all_seeds[n * p.batch_size:(n + 1) * p.batch_size]
@@ -1201,6 +1210,13 @@ class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing):
                 with devices.autocast():
                     extra_networks.activate(self, self.extra_network_data)
 
+    def get_conds(self):
+        if self.is_hr_pass:
+            return self.hr_c, self.hr_uc
+
+        return super().get_conds()
+
+
     def parse_extra_network_prompts(self):
         res = super().parse_extra_network_prompts()
 

modules/sd_models.py

@@ -295,11 +295,27 @@ def get_checkpoint_state_dict(checkpoint_info: CheckpointInfo, timer):
     return res
 
 
+class SkipWritingToConfig:
+    """This context manager prevents load_model_weights from writing checkpoint name to the config when it loads weight."""
+
+    skip = False
+    previous = None
+
+    def __enter__(self):
+        self.previous = SkipWritingToConfig.skip
+        SkipWritingToConfig.skip = True
+        return self
+
+    def __exit__(self, exc_type, exc_value, exc_traceback):
+        SkipWritingToConfig.skip = self.previous
+
+
 def load_model_weights(model, checkpoint_info: CheckpointInfo, state_dict, timer):
     sd_model_hash = checkpoint_info.calculate_shorthash()
     timer.record("calculate hash")
 
-    shared.opts.data["sd_model_checkpoint"] = checkpoint_info.title
+    if not SkipWritingToConfig.skip:
+        shared.opts.data["sd_model_checkpoint"] = checkpoint_info.title
 
     if state_dict is None:
         state_dict = get_checkpoint_state_dict(checkpoint_info, timer)
@@ -624,8 +640,11 @@ def reuse_model_from_already_loaded(sd_model, checkpoint_info, timer):
         timer.record("send model to device")
 
         model_data.set_sd_model(already_loaded)
-        shared.opts.data["sd_model_checkpoint"] = already_loaded.sd_checkpoint_info.title
-        shared.opts.data["sd_checkpoint_hash"] = already_loaded.sd_checkpoint_info.sha256
+
+        if not SkipWritingToConfig.skip:
+            shared.opts.data["sd_model_checkpoint"] = already_loaded.sd_checkpoint_info.title
+            shared.opts.data["sd_checkpoint_hash"] = already_loaded.sd_checkpoint_info.sha256
+
         print(f"Using already loaded model {already_loaded.sd_checkpoint_info.title}: done in {timer.summary()}")
         return model_data.sd_model
     elif shared.opts.sd_checkpoints_limit > 1 and len(model_data.loaded_sd_models) < shared.opts.sd_checkpoints_limit:

modules/sd_samplers_cfg_denoiser.py

@@ -38,16 +38,24 @@ class CFGDenoiser(torch.nn.Module):
     negative prompt.
     """
 
-    def __init__(self, model, sampler):
+    def __init__(self, sampler):
         super().__init__()
-        self.inner_model = model
+        self.model_wrap = None
         self.mask = None
         self.nmask = None
         self.init_latent = None
+        self.steps = None
         self.step = 0
         self.image_cfg_scale = None
         self.padded_cond_uncond = False
         self.sampler = sampler
+        self.model_wrap = None
+        self.p = None
+
+    @property
+    def inner_model(self):
+        raise NotImplementedError()
+
 
     def combine_denoised(self, x_out, conds_list, uncond, cond_scale):
         denoised_uncond = x_out[-uncond.shape[0]:]
@@ -68,10 +76,21 @@ class CFGDenoiser(torch.nn.Module):
     def get_pred_x0(self, x_in, x_out, sigma):
         return x_out
 
+    def update_inner_model(self):
+        self.model_wrap = None
+
+        c, uc = self.p.get_conds()
+        self.sampler.sampler_extra_args['cond'] = c
+        self.sampler.sampler_extra_args['uncond'] = uc
+
     def forward(self, x, sigma, uncond, cond, cond_scale, s_min_uncond, image_cond):
         if state.interrupted or state.skipped:
             raise sd_samplers_common.InterruptedException
 
+        if sd_samplers_common.apply_refiner(self):
+            cond = self.sampler.sampler_extra_args['cond']
+            uncond = self.sampler.sampler_extra_args['uncond']
+
         # at self.image_cfg_scale == 1.0 produced results for edit model are the same as with normal sampling,
         # so is_edit_model is set to False to support AND composition.
         is_edit_model = shared.sd_model.cond_stage_key == "edit" and self.image_cfg_scale is not None and self.image_cfg_scale != 1.0

modules/sd_samplers_common.py

@@ -3,7 +3,7 @@ from collections import namedtuple
 import numpy as np
 import torch
 from PIL import Image
-from modules import devices, images, sd_vae_approx, sd_samplers, sd_vae_taesd, shared
+from modules import devices, images, sd_vae_approx, sd_samplers, sd_vae_taesd, shared, sd_models
 from modules.shared import opts, state
 import k_diffusion.sampling
 
@@ -131,6 +131,35 @@ def replace_torchsde_browinan():
 replace_torchsde_browinan()
 
 
+def apply_refiner(sampler):
+    completed_ratio = sampler.step / sampler.steps
+
+    if completed_ratio <= shared.opts.sd_refiner_switch_at:
+        return False
+
+    if shared.opts.sd_refiner_checkpoint == "None":
+        return False
+
+    if shared.sd_model.sd_checkpoint_info.title == shared.opts.sd_refiner_checkpoint:
+        return False
+
+    refiner_checkpoint_info = sd_models.get_closet_checkpoint_match(shared.opts.sd_refiner_checkpoint)
+    if refiner_checkpoint_info is None:
+        raise Exception(f'Could not find checkpoint with name {shared.opts.sd_refiner_checkpoint}')
+
+    sampler.p.extra_generation_params['Refiner'] = refiner_checkpoint_info.short_title
+    sampler.p.extra_generation_params['Refiner switch at'] = shared.opts.sd_refiner_switch_at
+
+    with sd_models.SkipWritingToConfig():
+        sd_models.reload_model_weights(info=refiner_checkpoint_info)
+
+    devices.torch_gc()
+    sampler.p.setup_conds()
+    sampler.update_inner_model()
+
+    return True
+
+
 class TorchHijack:
     """This is here to replace torch.randn_like of k-diffusion.
 
@@ -176,8 +205,9 @@ class Sampler:
 
         self.conditioning_key = shared.sd_model.model.conditioning_key
 
-        self.model_wrap = None
+        self.p = None
         self.model_wrap_cfg = None
+        self.sampler_extra_args = None
 
     def callback_state(self, d):
         step = d['i']
@@ -189,6 +219,7 @@ class Sampler:
         shared.total_tqdm.update()
 
     def launch_sampling(self, steps, func):
+        self.model_wrap_cfg.steps = steps
         state.sampling_steps = steps
         state.sampling_step = 0
 
@@ -208,6 +239,8 @@ class Sampler:
         return p.steps
 
     def initialize(self, p) -> dict:
+        self.p = p
+        self.model_wrap_cfg.p = p
         self.model_wrap_cfg.mask = p.mask if hasattr(p, 'mask') else None
         self.model_wrap_cfg.nmask = p.nmask if hasattr(p, 'nmask') else None
         self.model_wrap_cfg.step = 0

modules/sd_samplers_kdiffusion.py

@@ -1,8 +1,7 @@
 import torch
 import inspect
 import k_diffusion.sampling
-from modules import sd_samplers_common, sd_samplers_extra
-from modules.sd_samplers_cfg_denoiser import CFGDenoiser
+from modules import sd_samplers_common, sd_samplers_extra, sd_samplers_cfg_denoiser
 
 from modules.shared import opts
 import modules.shared as shared
@@ -53,17 +52,24 @@ k_diffusion_scheduler = {
 }
 
 
+class CFGDenoiserKDiffusion(sd_samplers_cfg_denoiser.CFGDenoiser):
+    @property
+    def inner_model(self):
+        if self.model_wrap is None:
+            denoiser = k_diffusion.external.CompVisVDenoiser if shared.sd_model.parameterization == "v" else k_diffusion.external.CompVisDenoiser
+            self.model_wrap = denoiser(shared.sd_model, quantize=shared.opts.enable_quantization)
+
+        return self.model_wrap
+
+
 class KDiffusionSampler(sd_samplers_common.Sampler):
     def __init__(self, funcname, sd_model):
-
         super().__init__(funcname)
 
-        self.extra_params = sampler_extra_params.get(funcname, [])
         self.func = funcname if callable(funcname) else getattr(k_diffusion.sampling, self.funcname)
 
-        denoiser = k_diffusion.external.CompVisVDenoiser if sd_model.parameterization == "v" else k_diffusion.external.CompVisDenoiser
-        self.model_wrap = denoiser(sd_model, quantize=shared.opts.enable_quantization)
-        self.model_wrap_cfg = CFGDenoiser(self.model_wrap, self)
+        self.model_wrap_cfg = CFGDenoiserKDiffusion(self)
+        self.model_wrap = self.model_wrap_cfg.inner_model
 
     def get_sigmas(self, p, steps):
         discard_next_to_last_sigma = self.config is not None and self.config.options.get('discard_next_to_last_sigma', False)
@@ -144,7 +150,7 @@ class KDiffusionSampler(sd_samplers_common.Sampler):
 
         self.model_wrap_cfg.init_latent = x
         self.last_latent = x
-        extra_args = {
+        self.sampler_extra_args = {
             'cond': conditioning,
             'image_cond': image_conditioning,
             'uncond': unconditional_conditioning,
@@ -152,7 +158,7 @@ class KDiffusionSampler(sd_samplers_common.Sampler):
             's_min_uncond': self.s_min_uncond
         }
 
-        samples = self.launch_sampling(t_enc + 1, lambda: self.func(self.model_wrap_cfg, xi, extra_args=extra_args, disable=False, callback=self.callback_state, **extra_params_kwargs))
+        samples = self.launch_sampling(t_enc + 1, lambda: self.func(self.model_wrap_cfg, xi, extra_args=self.sampler_extra_args, disable=False, callback=self.callback_state, **extra_params_kwargs))
 
         if self.model_wrap_cfg.padded_cond_uncond:
             p.extra_generation_params["Pad conds"] = True
@@ -184,13 +190,14 @@ class KDiffusionSampler(sd_samplers_common.Sampler):
             extra_params_kwargs['noise_sampler'] = noise_sampler
 
         self.last_latent = x
-        samples = self.launch_sampling(steps, lambda: self.func(self.model_wrap_cfg, x, extra_args={
+        self.sampler_extra_args = {
             'cond': conditioning,
             'image_cond': image_conditioning,
             'uncond': unconditional_conditioning,
             'cond_scale': p.cfg_scale,
             's_min_uncond': self.s_min_uncond
-        }, disable=False, callback=self.callback_state, **extra_params_kwargs))
+        }
+        samples = self.launch_sampling(steps, lambda: self.func(self.model_wrap_cfg, x, extra_args=self.sampler_extra_args, disable=False, callback=self.callback_state, **extra_params_kwargs))
 
         if self.model_wrap_cfg.padded_cond_uncond:
             p.extra_generation_params["Pad conds"] = True

modules/sd_samplers_timesteps.py

@@ -45,10 +45,10 @@ class CompVisTimestepsVDenoiser(torch.nn.Module):
 
 class CFGDenoiserTimesteps(CFGDenoiser):
 
-    def __init__(self, model, sampler):
-        super().__init__(model, sampler)
+    def __init__(self, sampler):
+        super().__init__(sampler)
 
-        self.alphas = model.inner_model.alphas_cumprod
+        self.alphas = shared.sd_model.alphas_cumprod
 
     def get_pred_x0(self, x_in, x_out, sigma):
         ts = int(sigma.item())
@@ -61,6 +61,14 @@ class CFGDenoiserTimesteps(CFGDenoiser):
 
         return pred_x0
 
+    @property
+    def inner_model(self):
+        if self.model_wrap is None:
+            denoiser = CompVisTimestepsVDenoiser if shared.sd_model.parameterization == "v" else CompVisTimestepsDenoiser
+            self.model_wrap = denoiser(shared.sd_model)
+
+        return self.model_wrap
+
 
 class CompVisSampler(sd_samplers_common.Sampler):
     def __init__(self, funcname, sd_model):
@@ -69,9 +77,7 @@ class CompVisSampler(sd_samplers_common.Sampler):
         self.eta_option_field = 'eta_ddim'
         self.eta_infotext_field = 'Eta DDIM'
 
-        denoiser = CompVisTimestepsVDenoiser if sd_model.parameterization == "v" else CompVisTimestepsDenoiser
-        self.model_wrap = denoiser(sd_model)
-        self.model_wrap_cfg = CFGDenoiserTimesteps(self.model_wrap, self)
+        self.model_wrap_cfg = CFGDenoiserTimesteps(self)
 
     def get_timesteps(self, p, steps):
         discard_next_to_last_sigma = self.config is not None and self.config.options.get('discard_next_to_last_sigma', False)
@@ -107,7 +113,7 @@ class CompVisSampler(sd_samplers_common.Sampler):
 
         self.model_wrap_cfg.init_latent = x
         self.last_latent = x
-        extra_args = {
+        self.sampler_extra_args = {
             'cond': conditioning,
             'image_cond': image_conditioning,
             'uncond': unconditional_conditioning,
@@ -115,7 +121,7 @@ class CompVisSampler(sd_samplers_common.Sampler):
             's_min_uncond': self.s_min_uncond
         }
 
-        samples = self.launch_sampling(t_enc + 1, lambda: self.func(self.model_wrap_cfg, xi, extra_args=extra_args, disable=False, callback=self.callback_state, **extra_params_kwargs))
+        samples = self.launch_sampling(t_enc + 1, lambda: self.func(self.model_wrap_cfg, xi, extra_args=self.sampler_extra_args, disable=False, callback=self.callback_state, **extra_params_kwargs))
 
         if self.model_wrap_cfg.padded_cond_uncond:
             p.extra_generation_params["Pad conds"] = True
@@ -133,13 +139,14 @@ class CompVisSampler(sd_samplers_common.Sampler):
             extra_params_kwargs['timesteps'] = timesteps
 
         self.last_latent = x
-        samples = self.launch_sampling(steps, lambda: self.func(self.model_wrap_cfg, x, extra_args={
+        self.sampler_extra_args = {
             'cond': conditioning,
             'image_cond': image_conditioning,
             'uncond': unconditional_conditioning,
             'cond_scale': p.cfg_scale,
             's_min_uncond': self.s_min_uncond
-        }, disable=False, callback=self.callback_state, **extra_params_kwargs))
+        }
+        samples = self.launch_sampling(steps, lambda: self.func(self.model_wrap_cfg, x, extra_args=self.sampler_extra_args, disable=False, callback=self.callback_state, **extra_params_kwargs))
 
         if self.model_wrap_cfg.padded_cond_uncond:
             p.extra_generation_params["Pad conds"] = True

modules/shared_options.py

@@ -140,6 +140,8 @@ options_templates.update(options_section(('sd', "Stable Diffusion"), {
     "upcast_attn": OptionInfo(False, "Upcast cross attention layer to float32"),
     "randn_source": OptionInfo("GPU", "Random number generator source.", gr.Radio, {"choices": ["GPU", "CPU", "NV"]}).info("changes seeds drastically; use CPU to produce the same picture across different videocard vendors; use NV to produce same picture as on NVidia videocards"),
     "tiling": OptionInfo(False, "Tiling", infotext='Tiling').info("produce a tileable picture"),
+    "sd_refiner_checkpoint": OptionInfo("None", "Refiner checkpoint", gr.Dropdown, lambda: {"choices": ["None"] + shared_items.list_checkpoint_tiles()}, refresh=shared_items.refresh_checkpoints, infotext="Refiner").info("switch to another model in the middle of generation"),
+    "sd_refiner_switch_at": OptionInfo(1.0, "Refiner switch at", gr.Slider, {"minimum": 0.01, "maximum": 1.0, "step": 0.01}, infotext='Refiner switch at').info("fraction of sampling steps when the swtch to refiner model should happen; 1=never, 0.5=switch in the middle of generation"),
 }))
 
 options_templates.update(options_section(('sdxl', "Stable Diffusion XL"), {