From 8a34671fe91e142bce9e5556cca2258b3be9dd6e Mon Sep 17 00:00:00 2001 From: MrCheeze Date: Fri, 24 Mar 2023 22:48:16 -0400 Subject: Add support for the Variations models (unclip-h and unclip-l) --- modules/sd_samplers_kdiffusion.py | 19 +++++++++++++------ 1 file changed, 13 insertions(+), 6 deletions(-) (limited to 'modules/sd_samplers_kdiffusion.py') diff --git a/modules/sd_samplers_kdiffusion.py b/modules/sd_samplers_kdiffusion.py index 93f0e55a..e9f08518 100644 --- a/modules/sd_samplers_kdiffusion.py +++ b/modules/sd_samplers_kdiffusion.py @@ -92,14 +92,21 @@ class CFGDenoiser(torch.nn.Module): batch_size = len(conds_list) repeats = [len(conds_list[i]) for i in range(batch_size)] + if shared.sd_model.model.conditioning_key == "crossattn-adm": + image_uncond = torch.zeros_like(image_cond) + make_condition_dict = lambda c_crossattn, c_adm: {"c_crossattn": c_crossattn, "c_adm": c_adm} + else: + image_uncond = image_cond + make_condition_dict = lambda c_crossattn, c_concat: {"c_crossattn": c_crossattn, "c_concat": [c_concat]} + if not is_edit_model: x_in = torch.cat([torch.stack([x[i] for _ in range(n)]) for i, n in enumerate(repeats)] + [x]) sigma_in = torch.cat([torch.stack([sigma[i] for _ in range(n)]) for i, n in enumerate(repeats)] + [sigma]) - image_cond_in = torch.cat([torch.stack([image_cond[i] for _ in range(n)]) for i, n in enumerate(repeats)] + [image_cond]) + image_cond_in = torch.cat([torch.stack([image_cond[i] for _ in range(n)]) for i, n in enumerate(repeats)] + [image_uncond]) else: x_in = torch.cat([torch.stack([x[i] for _ in range(n)]) for i, n in enumerate(repeats)] + [x] + [x]) sigma_in = torch.cat([torch.stack([sigma[i] for _ in range(n)]) for i, n in enumerate(repeats)] + [sigma] + [sigma]) - image_cond_in = torch.cat([torch.stack([image_cond[i] for _ in range(n)]) for i, n in enumerate(repeats)] + [image_cond] + [torch.zeros_like(self.init_latent)]) + image_cond_in = torch.cat([torch.stack([image_cond[i] for _ in range(n)]) for i, n in enumerate(repeats)] + [image_uncond] + [torch.zeros_like(self.init_latent)]) denoiser_params = CFGDenoiserParams(x_in, image_cond_in, sigma_in, state.sampling_step, state.sampling_steps, tensor, uncond) cfg_denoiser_callback(denoiser_params) @@ -116,13 +123,13 @@ class CFGDenoiser(torch.nn.Module): cond_in = torch.cat([tensor, uncond, uncond]) if shared.batch_cond_uncond: - x_out = self.inner_model(x_in, sigma_in, cond={"c_crossattn": [cond_in], "c_concat": [image_cond_in]}) + x_out = self.inner_model(x_in, sigma_in, cond=make_condition_dict([cond_in], image_cond_in)) else: x_out = torch.zeros_like(x_in) for batch_offset in range(0, x_out.shape[0], batch_size): a = batch_offset b = a + batch_size - x_out[a:b] = self.inner_model(x_in[a:b], sigma_in[a:b], cond={"c_crossattn": [cond_in[a:b]], "c_concat": [image_cond_in[a:b]]}) + x_out[a:b] = self.inner_model(x_in[a:b], sigma_in[a:b], cond=make_condition_dict([cond_in[a:b]], image_cond_in[a:b])) else: x_out = torch.zeros_like(x_in) batch_size = batch_size*2 if shared.batch_cond_uncond else batch_size @@ -135,9 +142,9 @@ class CFGDenoiser(torch.nn.Module): else: c_crossattn = torch.cat([tensor[a:b]], uncond) - x_out[a:b] = self.inner_model(x_in[a:b], sigma_in[a:b], cond={"c_crossattn": c_crossattn, "c_concat": [image_cond_in[a:b]]}) + x_out[a:b] = self.inner_model(x_in[a:b], sigma_in[a:b], cond=make_condition_dict(c_crossattn, image_cond_in[a:b])) - x_out[-uncond.shape[0]:] = self.inner_model(x_in[-uncond.shape[0]:], sigma_in[-uncond.shape[0]:], cond={"c_crossattn": [uncond], "c_concat": [image_cond_in[-uncond.shape[0]:]]}) + x_out[-uncond.shape[0]:] = self.inner_model(x_in[-uncond.shape[0]:], sigma_in[-uncond.shape[0]:], cond=make_condition_dict([uncond], image_cond_in[-uncond.shape[0]:])) denoised_params = CFGDenoisedParams(x_out, state.sampling_step, state.sampling_steps) cfg_denoised_callback(denoised_params) -- cgit v1.2.3 From 42082e8a3239c1c32cd9e2a03a20b610af857b51 Mon Sep 17 00:00:00 2001 From: devdn Date: Tue, 28 Mar 2023 18:18:28 -0400 Subject: performance increase --- modules/sd_samplers_kdiffusion.py | 22 +++++++++++++++++----- 1 file changed, 17 insertions(+), 5 deletions(-) (limited to 'modules/sd_samplers_kdiffusion.py') diff --git a/modules/sd_samplers_kdiffusion.py b/modules/sd_samplers_kdiffusion.py index e9f08518..6a54ce32 100644 --- a/modules/sd_samplers_kdiffusion.py +++ b/modules/sd_samplers_kdiffusion.py @@ -76,7 +76,7 @@ class CFGDenoiser(torch.nn.Module): return denoised - def forward(self, x, sigma, uncond, cond, cond_scale, image_cond): + 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 @@ -116,6 +116,12 @@ class CFGDenoiser(torch.nn.Module): tensor = denoiser_params.text_cond uncond = denoiser_params.text_uncond + sigma_thresh = s_min_uncond + if(torch.dot(sigma,sigma) < sigma.shape[0] * (sigma_thresh*sigma_thresh) and not is_edit_model): + uncond = torch.zeros([0,0,uncond.shape[2]]) + x_in=x_in[:x_in.shape[0]//2] + sigma_in=sigma_in[:sigma_in.shape[0]//2] + if tensor.shape[1] == uncond.shape[1]: if not is_edit_model: cond_in = torch.cat([tensor, uncond]) @@ -144,7 +150,8 @@ class CFGDenoiser(torch.nn.Module): x_out[a:b] = self.inner_model(x_in[a:b], sigma_in[a:b], cond=make_condition_dict(c_crossattn, image_cond_in[a:b])) - x_out[-uncond.shape[0]:] = self.inner_model(x_in[-uncond.shape[0]:], sigma_in[-uncond.shape[0]:], cond=make_condition_dict([uncond], image_cond_in[-uncond.shape[0]:])) + if uncond.shape[0]: + x_out[-uncond.shape[0]:] = self.inner_model(x_in[-uncond.shape[0]:], sigma_in[-uncond.shape[0]:], cond=make_condition_dict([uncond], image_cond_in[-uncond.shape[0]:])) denoised_params = CFGDenoisedParams(x_out, state.sampling_step, state.sampling_steps) cfg_denoised_callback(denoised_params) @@ -157,7 +164,10 @@ class CFGDenoiser(torch.nn.Module): sd_samplers_common.store_latent(x_out[-uncond.shape[0]:]) if not is_edit_model: - denoised = self.combine_denoised(x_out, conds_list, uncond, cond_scale) + if uncond.shape[0]: + denoised = self.combine_denoised(x_out, conds_list, uncond, cond_scale) + else: + denoised = x_out else: denoised = self.combine_denoised_for_edit_model(x_out, cond_scale) @@ -165,7 +175,6 @@ class CFGDenoiser(torch.nn.Module): denoised = self.init_latent * self.mask + self.nmask * denoised self.step += 1 - return denoised @@ -244,6 +253,7 @@ class KDiffusionSampler: self.model_wrap_cfg.step = 0 self.model_wrap_cfg.image_cfg_scale = getattr(p, 'image_cfg_scale', None) self.eta = p.eta if p.eta is not None else opts.eta_ancestral + self.s_min_uncond = getattr(p, 's_min_uncond', 0.0) k_diffusion.sampling.torch = TorchHijack(self.sampler_noises if self.sampler_noises is not None else []) @@ -326,6 +336,7 @@ class KDiffusionSampler: 'image_cond': image_conditioning, 'uncond': unconditional_conditioning, 'cond_scale': p.cfg_scale, + '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)) @@ -359,7 +370,8 @@ class KDiffusionSampler: 'cond': conditioning, 'image_cond': image_conditioning, 'uncond': unconditional_conditioning, - 'cond_scale': p.cfg_scale + 'cond_scale': p.cfg_scale, + 's_min_uncond': self.s_min_uncond }, disable=False, callback=self.callback_state, **extra_params_kwargs)) return samples -- cgit v1.2.3 From 44e8e9c36807d4a71c2fc84129ebcf5ba4f77f21 Mon Sep 17 00:00:00 2001 From: devdn Date: Thu, 30 Mar 2023 00:54:28 -0400 Subject: fix live preview & alternate uncond guidance for better quality --- modules/sd_samplers_kdiffusion.py | 14 ++++++++------ 1 file changed, 8 insertions(+), 6 deletions(-) (limited to 'modules/sd_samplers_kdiffusion.py') diff --git a/modules/sd_samplers_kdiffusion.py b/modules/sd_samplers_kdiffusion.py index 6a54ce32..17d24df4 100644 --- a/modules/sd_samplers_kdiffusion.py +++ b/modules/sd_samplers_kdiffusion.py @@ -116,11 +116,13 @@ class CFGDenoiser(torch.nn.Module): tensor = denoiser_params.text_cond uncond = denoiser_params.text_uncond - sigma_thresh = s_min_uncond - if(torch.dot(sigma,sigma) < sigma.shape[0] * (sigma_thresh*sigma_thresh) and not is_edit_model): - uncond = torch.zeros([0,0,uncond.shape[2]]) - x_in=x_in[:x_in.shape[0]//2] - sigma_in=sigma_in[:sigma_in.shape[0]//2] + if self.step % 2 and s_min_uncond > 0 and not is_edit_model: + # alternating uncond allows for higher thresholds without the quality loss normally expected from raising it + sigma_threshold = s_min_uncond + if(torch.dot(sigma,sigma) < sigma.shape[0] * (sigma_threshold*sigma_threshold) ): + uncond = torch.zeros([0,0,uncond.shape[2]]) + x_in=x_in[:x_in.shape[0]//2] + sigma_in=sigma_in[:sigma_in.shape[0]//2] if tensor.shape[1] == uncond.shape[1]: if not is_edit_model: @@ -159,7 +161,7 @@ class CFGDenoiser(torch.nn.Module): devices.test_for_nans(x_out, "unet") if opts.live_preview_content == "Prompt": - sd_samplers_common.store_latent(x_out[0:uncond.shape[0]]) + sd_samplers_common.store_latent(x_out[0:x_out.shape[0]-uncond.shape[0]]) elif opts.live_preview_content == "Negative prompt": sd_samplers_common.store_latent(x_out[-uncond.shape[0]:]) -- cgit v1.2.3 From d40e44ade479f7bba30d5317381cbc58c861775b Mon Sep 17 00:00:00 2001 From: Deciare <1689220+deciare@users.noreply.github.com> Date: Tue, 18 Apr 2023 23:18:58 -0400 Subject: Option to use CPU for random number generation. Makes a given manual seed generate the same images across different platforms, independently of the GPU architecture in use. Fixes #9613. --- modules/sd_samplers_kdiffusion.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules/sd_samplers_kdiffusion.py') diff --git a/modules/sd_samplers_kdiffusion.py b/modules/sd_samplers_kdiffusion.py index e9f08518..13f4567a 100644 --- a/modules/sd_samplers_kdiffusion.py +++ b/modules/sd_samplers_kdiffusion.py @@ -190,7 +190,7 @@ class TorchHijack: if noise.shape == x.shape: return noise - if x.device.type == 'mps': + if opts.use_cpu_randn or x.device.type == 'mps': return torch.randn_like(x, device=devices.cpu).to(x.device) else: return torch.randn_like(x) -- cgit v1.2.3 From 5fe0dd79beaa5ef737ff85254ee9870f60ae9464 Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Sat, 29 Apr 2023 11:29:37 +0300 Subject: rename CPU RNG to RNG source in settings, add infotext and parameters copypaste support to RNG source --- modules/sd_samplers_kdiffusion.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules/sd_samplers_kdiffusion.py') diff --git a/modules/sd_samplers_kdiffusion.py b/modules/sd_samplers_kdiffusion.py index 13f4567a..a547d1b5 100644 --- a/modules/sd_samplers_kdiffusion.py +++ b/modules/sd_samplers_kdiffusion.py @@ -190,7 +190,7 @@ class TorchHijack: if noise.shape == x.shape: return noise - if opts.use_cpu_randn or x.device.type == 'mps': + if opts.randn_source == "CPU" or x.device.type == 'mps': return torch.randn_like(x, device=devices.cpu).to(x.device) else: return torch.randn_like(x) -- cgit v1.2.3 From 1d11e896984c883f6a0debb3abaef945595cbc70 Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Sat, 29 Apr 2023 15:57:09 +0300 Subject: rework Negative Guidance minimum sigma to work with AND, add infotext and copypaste parameters support --- modules/sd_samplers_kdiffusion.py | 43 +++++++++++++++++++++++---------------- 1 file changed, 25 insertions(+), 18 deletions(-) (limited to 'modules/sd_samplers_kdiffusion.py') diff --git a/modules/sd_samplers_kdiffusion.py b/modules/sd_samplers_kdiffusion.py index d42d5fcf..f8aaac59 100644 --- a/modules/sd_samplers_kdiffusion.py +++ b/modules/sd_samplers_kdiffusion.py @@ -115,20 +115,21 @@ class CFGDenoiser(torch.nn.Module): sigma_in = denoiser_params.sigma tensor = denoiser_params.text_cond uncond = denoiser_params.text_uncond + skip_uncond = False - if self.step % 2 and s_min_uncond > 0 and not is_edit_model: - # alternating uncond allows for higher thresholds without the quality loss normally expected from raising it - sigma_threshold = s_min_uncond - if(torch.dot(sigma,sigma) < sigma.shape[0] * (sigma_threshold*sigma_threshold) ): - uncond = torch.zeros([0,0,uncond.shape[2]]) - x_in=x_in[:x_in.shape[0]//2] - sigma_in=sigma_in[:sigma_in.shape[0]//2] + # alternating uncond allows for higher thresholds without the quality loss normally expected from raising it + if self.step % 2 and s_min_uncond > 0 and sigma[0] < s_min_uncond and not is_edit_model: + skip_uncond = True + x_in = x_in[:-batch_size] + sigma_in = sigma_in[:-batch_size] - if tensor.shape[1] == uncond.shape[1]: - if not is_edit_model: - cond_in = torch.cat([tensor, uncond]) - else: + if tensor.shape[1] == uncond.shape[1] or skip_uncond: + if is_edit_model: cond_in = torch.cat([tensor, uncond, uncond]) + elif skip_uncond: + cond_in = tensor + else: + cond_in = torch.cat([tensor, uncond]) if shared.batch_cond_uncond: x_out = self.inner_model(x_in, sigma_in, cond=make_condition_dict([cond_in], image_cond_in)) @@ -152,9 +153,15 @@ class CFGDenoiser(torch.nn.Module): x_out[a:b] = self.inner_model(x_in[a:b], sigma_in[a:b], cond=make_condition_dict(c_crossattn, image_cond_in[a:b])) - if uncond.shape[0]: + if not skip_uncond: x_out[-uncond.shape[0]:] = self.inner_model(x_in[-uncond.shape[0]:], sigma_in[-uncond.shape[0]:], cond=make_condition_dict([uncond], image_cond_in[-uncond.shape[0]:])) + if skip_uncond: + #x_out = torch.cat([x_out, x_out[0:batch_size]]) # we skipped uncond denoising, so we put cond-denoised image to where the uncond-denoised image should be + denoised_image_indexes = [x[0][0] for x in conds_list] + fake_uncond = torch.cat([x_out[i:i+1] for i in denoised_image_indexes]) + x_out = torch.cat([x_out, fake_uncond]) + denoised_params = CFGDenoisedParams(x_out, state.sampling_step, state.sampling_steps) cfg_denoised_callback(denoised_params) @@ -165,13 +172,12 @@ class CFGDenoiser(torch.nn.Module): elif opts.live_preview_content == "Negative prompt": sd_samplers_common.store_latent(x_out[-uncond.shape[0]:]) - if not is_edit_model: - if uncond.shape[0]: - denoised = self.combine_denoised(x_out, conds_list, uncond, cond_scale) - else: - denoised = x_out - else: + if is_edit_model: denoised = self.combine_denoised_for_edit_model(x_out, cond_scale) + elif skip_uncond: + denoised = self.combine_denoised(x_out, conds_list, uncond, 1.0) + else: + denoised = self.combine_denoised(x_out, conds_list, uncond, cond_scale) if self.mask is not None: denoised = self.init_latent * self.mask + self.nmask * denoised @@ -221,6 +227,7 @@ class KDiffusionSampler: self.eta = None self.config = None self.last_latent = None + self.s_min_uncond = None self.conditioning_key = sd_model.model.conditioning_key -- cgit v1.2.3 From 737b73a820584b8035fcc37fe35993bec867f326 Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Sat, 29 Apr 2023 16:05:20 +0300 Subject: some extra lines I forgot to add for previous commit --- modules/sd_samplers_kdiffusion.py | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) (limited to 'modules/sd_samplers_kdiffusion.py') diff --git a/modules/sd_samplers_kdiffusion.py b/modules/sd_samplers_kdiffusion.py index f8aaac59..136aa8e5 100644 --- a/modules/sd_samplers_kdiffusion.py +++ b/modules/sd_samplers_kdiffusion.py @@ -156,11 +156,10 @@ class CFGDenoiser(torch.nn.Module): if not skip_uncond: x_out[-uncond.shape[0]:] = self.inner_model(x_in[-uncond.shape[0]:], sigma_in[-uncond.shape[0]:], cond=make_condition_dict([uncond], image_cond_in[-uncond.shape[0]:])) + denoised_image_indexes = [x[0][0] for x in conds_list] if skip_uncond: - #x_out = torch.cat([x_out, x_out[0:batch_size]]) # we skipped uncond denoising, so we put cond-denoised image to where the uncond-denoised image should be - denoised_image_indexes = [x[0][0] for x in conds_list] fake_uncond = torch.cat([x_out[i:i+1] for i in denoised_image_indexes]) - x_out = torch.cat([x_out, fake_uncond]) + x_out = torch.cat([x_out, fake_uncond]) # we skipped uncond denoising, so we put cond-denoised image to where the uncond-denoised image should be denoised_params = CFGDenoisedParams(x_out, state.sampling_step, state.sampling_steps) cfg_denoised_callback(denoised_params) -- cgit v1.2.3 From 8863b31d83b527d041ca45e23b9af99e2346081a Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Sat, 29 Apr 2023 16:06:20 +0300 Subject: use correct images for previews when using AND (see #9491) --- modules/sd_samplers_kdiffusion.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules/sd_samplers_kdiffusion.py') diff --git a/modules/sd_samplers_kdiffusion.py b/modules/sd_samplers_kdiffusion.py index 136aa8e5..eb98e599 100644 --- a/modules/sd_samplers_kdiffusion.py +++ b/modules/sd_samplers_kdiffusion.py @@ -167,7 +167,7 @@ class CFGDenoiser(torch.nn.Module): devices.test_for_nans(x_out, "unet") if opts.live_preview_content == "Prompt": - sd_samplers_common.store_latent(x_out[0:x_out.shape[0]-uncond.shape[0]]) + sd_samplers_common.store_latent(torch.cat([x_out[i:i+1] for i in denoised_image_indexes])) elif opts.live_preview_content == "Negative prompt": sd_samplers_common.store_latent(x_out[-uncond.shape[0]:]) -- cgit v1.2.3