From 9d40212485febe05a662dd0346e6def83e456288 Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Tue, 13 Sep 2022 21:49:58 +0300 Subject: first attempt to produce crrect seeds in batch --- modules/sd_samplers.py | 25 +++++++++++++++++++++++++ 1 file changed, 25 insertions(+) (limited to 'modules/sd_samplers.py') diff --git a/modules/sd_samplers.py b/modules/sd_samplers.py index 7ef507f1..f77fe43f 100644 --- a/modules/sd_samplers.py +++ b/modules/sd_samplers.py @@ -93,6 +93,10 @@ class VanillaStableDiffusionSampler: self.mask = None self.nmask = None self.init_latent = None + self.sampler_noises = None + + def number_of_needed_noises(self, p): + return 0 def sample_img2img(self, p, x, noise, conditioning, unconditional_conditioning): t_enc = int(min(p.denoising_strength, 0.999) * p.steps) @@ -171,16 +175,37 @@ def extended_trange(count, *args, **kwargs): shared.total_tqdm.update() +original_randn_like = torch.randn_like + class KDiffusionSampler: def __init__(self, funcname, sd_model): self.model_wrap = k_diffusion.external.CompVisDenoiser(sd_model) self.funcname = funcname self.func = getattr(k_diffusion.sampling, self.funcname) self.model_wrap_cfg = CFGDenoiser(self.model_wrap) + self.sampler_noises = None + self.sampler_noise_index = 0 + + k_diffusion.sampling.torch.randn_like = self.randn_like def callback_state(self, d): store_latent(d["denoised"]) + def number_of_needed_noises(self, p): + return p.steps + + def randn_like(self, x): + noise = self.sampler_noises[self.sampler_noise_index] if self.sampler_noises is not None and self.sampler_noise_index < len(self.sampler_noises) else None + + if noise is not None and x.shape == noise.shape: + res = noise + else: + print('generating') + res = original_randn_like(x) + + self.sampler_noise_index += 1 + return res + def sample_img2img(self, p, x, noise, conditioning, unconditional_conditioning): t_enc = int(min(p.denoising_strength, 0.999) * p.steps) sigmas = self.model_wrap.get_sigmas(p.steps) -- cgit v1.2.1 From b44ddcb44398fbe922fd7515f66d8b0c2344bc54 Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Fri, 16 Sep 2022 08:51:21 +0300 Subject: Prompt editing only applies to images in first batch of desired batch size when batch count > 1 #535 --- modules/sd_samplers.py | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'modules/sd_samplers.py') diff --git a/modules/sd_samplers.py b/modules/sd_samplers.py index df3a6fe8..5d95bfe0 100644 --- a/modules/sd_samplers.py +++ b/modules/sd_samplers.py @@ -115,6 +115,7 @@ class VanillaStableDiffusionSampler: self.mask = p.mask self.nmask = p.nmask self.init_latent = p.init_latent + self.step = 0 samples = self.sampler.decode(x1, conditioning, t_enc, unconditional_guidance_scale=p.cfg_scale, unconditional_conditioning=unconditional_conditioning) @@ -127,6 +128,7 @@ class VanillaStableDiffusionSampler: self.mask = None self.nmask = None self.init_latent = None + self.step = 0 # existing code fails with cetin step counts, like 9 try: @@ -206,6 +208,7 @@ class KDiffusionSampler: self.model_wrap_cfg.mask = p.mask self.model_wrap_cfg.nmask = p.nmask self.model_wrap_cfg.init_latent = p.init_latent + self.model_wrap.step = 0 if hasattr(k_diffusion.sampling, 'trange'): k_diffusion.sampling.trange = lambda *args, **kwargs: extended_trange(*args, **kwargs) @@ -216,6 +219,8 @@ class KDiffusionSampler: sigmas = self.model_wrap.get_sigmas(p.steps) x = x * sigmas[0] + self.model_wrap_cfg.step = 0 + if hasattr(k_diffusion.sampling, 'trange'): k_diffusion.sampling.trange = lambda *args, **kwargs: extended_trange(*args, **kwargs) -- cgit v1.2.1 From 87e8b9a2ab3f033e7fdadbb2fe258857915980ac Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Fri, 16 Sep 2022 09:47:03 +0300 Subject: prevent replacing torch_randn globally (instead replacing k_diffusion.sampling.torch) and add a setting to disable this all --- modules/sd_samplers.py | 25 ++++++++++++++++++++----- 1 file changed, 20 insertions(+), 5 deletions(-) (limited to 'modules/sd_samplers.py') diff --git a/modules/sd_samplers.py b/modules/sd_samplers.py index f77fe43f..d478c5bc 100644 --- a/modules/sd_samplers.py +++ b/modules/sd_samplers.py @@ -175,7 +175,19 @@ def extended_trange(count, *args, **kwargs): shared.total_tqdm.update() -original_randn_like = torch.randn_like +class TorchHijack: + def __init__(self, kdiff_sampler): + self.kdiff_sampler = kdiff_sampler + + def __getattr__(self, item): + if item == 'randn_like': + return self.kdiff_sampler.randn_like + + if hasattr(torch, item): + return getattr(torch, item) + + raise AttributeError("'{}' object has no attribute '{}'".format(type(self).__name__, item)) + class KDiffusionSampler: def __init__(self, funcname, sd_model): @@ -186,8 +198,6 @@ class KDiffusionSampler: self.sampler_noises = None self.sampler_noise_index = 0 - k_diffusion.sampling.torch.randn_like = self.randn_like - def callback_state(self, d): store_latent(d["denoised"]) @@ -200,8 +210,7 @@ class KDiffusionSampler: if noise is not None and x.shape == noise.shape: res = noise else: - print('generating') - res = original_randn_like(x) + res = torch.randn_like(x) self.sampler_noise_index += 1 return res @@ -223,6 +232,9 @@ class KDiffusionSampler: if hasattr(k_diffusion.sampling, 'trange'): k_diffusion.sampling.trange = lambda *args, **kwargs: extended_trange(*args, **kwargs) + if self.sampler_noises is not None: + k_diffusion.sampling.torch = TorchHijack(self) + return self.func(self.model_wrap_cfg, xi, sigma_sched, extra_args={'cond': conditioning, 'uncond': unconditional_conditioning, 'cond_scale': p.cfg_scale}, disable=False, callback=self.callback_state) def sample(self, p, x, conditioning, unconditional_conditioning): @@ -232,6 +244,9 @@ class KDiffusionSampler: if hasattr(k_diffusion.sampling, 'trange'): k_diffusion.sampling.trange = lambda *args, **kwargs: extended_trange(*args, **kwargs) + if self.sampler_noises is not None: + k_diffusion.sampling.torch = TorchHijack(self) + samples_ddim = self.func(self.model_wrap_cfg, x, sigmas, extra_args={'cond': conditioning, 'uncond': unconditional_conditioning, 'cond_scale': p.cfg_scale}, disable=False, callback=self.callback_state) return samples_ddim -- cgit v1.2.1 From e49b1c5d73ede818adb624590934f051b94493ac Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Fri, 16 Sep 2022 13:38:02 +0300 Subject: an option to do exactly the amount of specified steps in img2img --- modules/sd_samplers.py | 26 +++++++++++++++++++------- 1 file changed, 19 insertions(+), 7 deletions(-) (limited to 'modules/sd_samplers.py') diff --git a/modules/sd_samplers.py b/modules/sd_samplers.py index 02ffce0e..1b3dc302 100644 --- a/modules/sd_samplers.py +++ b/modules/sd_samplers.py @@ -38,6 +38,17 @@ samplers = [ samplers_for_img2img = [x for x in samplers if x.name != 'PLMS'] +def setup_img2img_steps(p): + if opts.img2img_fix_steps: + steps = int(p.steps / min(p.denoising_strength, 0.999)) + t_enc = p.steps - 1 + else: + steps = p.steps + t_enc = int(min(p.denoising_strength, 0.999) * steps) + + return steps, t_enc + + def sample_to_image(samples): x_sample = shared.sd_model.decode_first_stage(samples[0:1].type(shared.sd_model.dtype))[0] x_sample = torch.clamp((x_sample + 1.0) / 2.0, min=0.0, max=1.0) @@ -105,13 +116,13 @@ class VanillaStableDiffusionSampler: return res def sample_img2img(self, p, x, noise, conditioning, unconditional_conditioning): - t_enc = int(min(p.denoising_strength, 0.999) * p.steps) + steps, t_enc = setup_img2img_steps(p) # existing code fails with cetain step counts, like 9 try: - self.sampler.make_schedule(ddim_num_steps=p.steps, verbose=False) + self.sampler.make_schedule(ddim_num_steps=steps, verbose=False) except Exception: - self.sampler.make_schedule(ddim_num_steps=p.steps+1, verbose=False) + self.sampler.make_schedule(ddim_num_steps=steps+1, verbose=False) x1 = self.sampler.stochastic_encode(x, torch.tensor([t_enc] * int(x.shape[0])).to(shared.device), noise=noise) @@ -230,14 +241,15 @@ class KDiffusionSampler: return res def sample_img2img(self, p, x, noise, conditioning, unconditional_conditioning): - t_enc = int(min(p.denoising_strength, 0.999) * p.steps) - sigmas = self.model_wrap.get_sigmas(p.steps) + steps, t_enc = setup_img2img_steps(p) + + sigmas = self.model_wrap.get_sigmas(steps) - noise = noise * sigmas[p.steps - t_enc - 1] + noise = noise * sigmas[steps - t_enc - 1] xi = x + noise - sigma_sched = sigmas[p.steps - t_enc - 1:] + sigma_sched = sigmas[steps - t_enc - 1:] self.model_wrap_cfg.mask = p.mask self.model_wrap_cfg.nmask = p.nmask -- cgit v1.2.1