Closed h3rmit-git closed 1 year ago
dvruette
commented
1 year ago I've been struggling getting xformers to work in the past, but I believe that this is an error I haven't seen before, I'll look into. In general, I recommend --opt-split-attention when using FABRIC, especially if using > 2 feedback images as even xformers struggles with VRAM in those cases. I'll also be looking into improving performance (speed, memory) using token mergin or similar in the near future.
dvruette
commented
1 year ago I seem to be unable to reproduce the error above. It seems to occur in the VAE forward pass while multi-diffusion is enabled, but it doesn't happen for me even with the respective plugin enabled. Could you share some generation parameters to help reproduce the issue?
h3rmit-git
commented
1 year ago Below is a stacktrace without Multidiffusion.
I believe the error may be related to the version of something.
Stability AI version: https://github.com/Stability-AI/stablediffusion/commit/cf1d67a6fd5ea1aa600c4df58e5b47da45f6bdbf (Mar 25, 2023)
COMMANDLINE_ARGS=--disable-safe-unpickle --medvram --xformers --no-half-vae
The option --opt-split-attention unfortunately makes me run out of VRAM even when not using FABRIC. I'm running on 4GB of VRAM.
Traceback (most recent call last):
File "E:\stable-diffusion-webui\modules\call_queue.py", line 57, in f
res = list(func(*args, **kwargs))
File "E:\stable-diffusion-webui\modules\call_queue.py", line 37, in f
res = func(*args, **kwargs)
File "E:\stable-diffusion-webui\modules\txt2img.py", line 57, in txt2img
processed = processing.process_images(p)
File "E:\stable-diffusion-webui\modules\processing.py", line 611, in process_images
res = process_images_inner(p)
File "E:\stable-diffusion-webui\extensions\sd-webui-controlnet\scripts\batch_hijack.py", line 42, in processing_process_images_hijack
return getattr(processing, '__controlnet_original_process_images_inner')(p, *args, **kwargs)
File "E:\stable-diffusion-webui\modules\processing.py", line 729, in process_images_inner
samples_ddim = p.sample(conditioning=p.c, unconditional_conditioning=p.uc, seeds=p.seeds, subseeds=p.subseeds, subseed_strength=p.subseed_strength, prompts=p.prompts)
File "E:\stable-diffusion-webui\extensions\sd-webui-fabric\scripts\marking.py", line 28, in process_sample
return process.sample_before_CN_hack(*args, **kwargs)
File "E:\stable-diffusion-webui\modules\processing.py", line 977, in sample
samples = self.sampler.sample(self, x, conditioning, unconditional_conditioning, image_conditioning=self.txt2img_image_conditioning(x))
File "E:\stable-diffusion-webui\modules\sd_samplers_kdiffusion.py", line 383, in sample
samples = self.launch_sampling(steps, lambda: self.func(self.model_wrap_cfg, x, extra_args={
File "E:\stable-diffusion-webui\modules\sd_samplers_kdiffusion.py", line 257, in launch_sampling
return func()
File "E:\stable-diffusion-webui\modules\sd_samplers_kdiffusion.py", line 383, in <lambda>
samples = self.launch_sampling(steps, lambda: self.func(self.model_wrap_cfg, x, extra_args={
File "E:\stable-diffusion-webui\venv\lib\site-packages\torch\utils\_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
File "E:\stable-diffusion-webui\repositories\k-diffusion\k_diffusion\sampling.py", line 594, in sample_dpmpp_2m
denoised = model(x, sigmas[i] * s_in, **extra_args)
File "E:\stable-diffusion-webui\venv\lib\site-packages\torch\nn\modules\module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "E:\stable-diffusion-webui\modules\sd_samplers_kdiffusion.py", line 143, in forward
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]))
File "E:\stable-diffusion-webui\venv\lib\site-packages\torch\nn\modules\module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "E:\stable-diffusion-webui\repositories\k-diffusion\k_diffusion\external.py", line 112, in forward
eps = self.get_eps(input * c_in, self.sigma_to_t(sigma), **kwargs)
File "E:\stable-diffusion-webui\repositories\k-diffusion\k_diffusion\external.py", line 138, in get_eps
return self.inner_model.apply_model(*args, **kwargs)
File "E:\stable-diffusion-webui\modules\sd_hijack_utils.py", line 17, in <lambda>
setattr(resolved_obj, func_path[-1], lambda *args, **kwargs: self(*args, **kwargs))
File "E:\stable-diffusion-webui\modules\sd_hijack_utils.py", line 28, in __call__
return self.__orig_func(*args, **kwargs)
File "E:\stable-diffusion-webui\repositories\stable-diffusion-stability-ai\ldm\models\diffusion\ddpm.py", line 858, in apply_model
x_recon = self.model(x_noisy, t, **cond)
File "E:\stable-diffusion-webui\venv\lib\site-packages\torch\nn\modules\module.py", line 1538, in _call_impl
result = forward_call(*args, **kwargs)
File "E:\stable-diffusion-webui\repositories\stable-diffusion-stability-ai\ldm\models\diffusion\ddpm.py", line 1335, in forward
out = self.diffusion_model(x, t, context=cc)
File "E:\stable-diffusion-webui\venv\lib\site-packages\torch\nn\modules\module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "E:\stable-diffusion-webui\extensions\sd-webui-fabric\scripts\patching.py", line 105, in new_forward
pos_latents, neg_latents = get_latents_from_params(p, params, w, h)
File "E:\stable-diffusion-webui\extensions\sd-webui-fabric\scripts\patching.py", line 46, in get_latents_from_params
params.pos_latents = get_latents(params.pos_images, params.pos_latents)
File "E:\stable-diffusion-webui\extensions\sd-webui-fabric\scripts\patching.py", line 36, in get_latents
return [encode_to_latent(p, img, w, h) for img in images]
File "E:\stable-diffusion-webui\extensions\sd-webui-fabric\scripts\patching.py", line 36, in <listcomp>
return [encode_to_latent(p, img, w, h) for img in images]
File "E:\stable-diffusion-webui\extensions\sd-webui-fabric\scripts\patching.py", line 24, in encode_to_latent
vae_output = p.sd_model.encode_first_stage(x)
File "E:\stable-diffusion-webui\modules\sd_hijack_utils.py", line 17, in <lambda>
setattr(resolved_obj, func_path[-1], lambda *args, **kwargs: self(*args, **kwargs))
File "E:\stable-diffusion-webui\modules\sd_hijack_utils.py", line 28, in __call__
return self.__orig_func(*args, **kwargs)
File "E:\stable-diffusion-webui\venv\lib\site-packages\torch\utils\_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
File "E:\stable-diffusion-webui\repositories\stable-diffusion-stability-ai\ldm\models\diffusion\ddpm.py", line 830, in encode_first_stage
return self.first_stage_model.encode(x)
File "E:\stable-diffusion-webui\modules\lowvram.py", line 48, in first_stage_model_encode_wrap
return first_stage_model_encode(x)
File "E:\stable-diffusion-webui\repositories\stable-diffusion-stability-ai\ldm\models\autoencoder.py", line 83, in encode
h = self.encoder(x)
File "E:\stable-diffusion-webui\venv\lib\site-packages\torch\nn\modules\module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "E:\stable-diffusion-webui\repositories\stable-diffusion-stability-ai\ldm\modules\diffusionmodules\model.py", line 536, in forward
h = self.mid.attn_1(h)
File "E:\stable-diffusion-webui\venv\lib\site-packages\torch\nn\modules\module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "E:\stable-diffusion-webui\repositories\stable-diffusion-stability-ai\ldm\modules\diffusionmodules\model.py", line 258, in forward
out = xformers.ops.memory_efficient_attention(q, k, v, attn_bias=None, op=self.attention_op)
File "E:\stable-diffusion-webui\extensions\sd-webui-fabric\scripts\weighted_attention.py", line 36, in patched_xformers_attn
bs, nq, nh, dh = q.shape # batch_size, num_queries, num_heads, dim_per_headI unfortunately still can't reproduce the issue on my end. I tried pushing a blind fix based on the error message, but it's pretty much a shot in the dark at this point. Could you maybe provide the specific steps that lead to the error? Does it happen immediately or only after a few generations?
h3rmit-git
commented
1 year ago The error happens immediately and always. The steps I am following are:
After your last two commits that attempted to fix the issue, I am getting some new errors:
First run stacktrace:
Traceback (most recent call last):
File "E:\stable-diffusion-webui\modules\call_queue.py", line 57, in f
res = list(func(*args, **kwargs))
File "E:\stable-diffusion-webui\modules\call_queue.py", line 37, in f
res = func(*args, **kwargs)
File "E:\stable-diffusion-webui\modules\txt2img.py", line 57, in txt2img
processed = processing.process_images(p)
File "E:\stable-diffusion-webui\modules\processing.py", line 611, in process_images
res = process_images_inner(p)
File "E:\stable-diffusion-webui\extensions\sd-webui-controlnet\scripts\batch_hijack.py", line 42, in processing_process_images_hijack
return getattr(processing, '__controlnet_original_process_images_inner')(p, *args, **kwargs)
File "E:\stable-diffusion-webui\modules\processing.py", line 729, in process_images_inner
samples_ddim = p.sample(conditioning=p.c, unconditional_conditioning=p.uc, seeds=p.seeds, subseeds=p.subseeds, subseed_strength=p.subseed_strength, prompts=p.prompts)
File "E:\stable-diffusion-webui\extensions\sd-webui-fabric\scripts\marking.py", line 28, in process_sample
return process.sample_before_CN_hack(*args, **kwargs)
File "E:\stable-diffusion-webui\modules\processing.py", line 977, in sample
samples = self.sampler.sample(self, x, conditioning, unconditional_conditioning, image_conditioning=self.txt2img_image_conditioning(x))
File "E:\stable-diffusion-webui\modules\sd_samplers_kdiffusion.py", line 383, in sample
samples = self.launch_sampling(steps, lambda: self.func(self.model_wrap_cfg, x, extra_args={
File "E:\stable-diffusion-webui\modules\sd_samplers_kdiffusion.py", line 257, in launch_sampling
return func()
File "E:\stable-diffusion-webui\modules\sd_samplers_kdiffusion.py", line 383, in <lambda>
samples = self.launch_sampling(steps, lambda: self.func(self.model_wrap_cfg, x, extra_args={
File "E:\stable-diffusion-webui\venv\lib\site-packages\torch\utils\_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
File "E:\stable-diffusion-webui\repositories\k-diffusion\k_diffusion\sampling.py", line 626, in sample_dpmpp_2m_sde
denoised = model(x, sigmas[i] * s_in, **extra_args)
File "E:\stable-diffusion-webui\venv\lib\site-packages\torch\nn\modules\module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "E:\stable-diffusion-webui\modules\sd_samplers_kdiffusion.py", line 143, in forward
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]))
File "E:\stable-diffusion-webui\venv\lib\site-packages\torch\nn\modules\module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "E:\stable-diffusion-webui\repositories\k-diffusion\k_diffusion\external.py", line 112, in forward
eps = self.get_eps(input * c_in, self.sigma_to_t(sigma), **kwargs)
File "E:\stable-diffusion-webui\repositories\k-diffusion\k_diffusion\external.py", line 138, in get_eps
return self.inner_model.apply_model(*args, **kwargs)
File "E:\stable-diffusion-webui\modules\sd_hijack_utils.py", line 17, in <lambda>
setattr(resolved_obj, func_path[-1], lambda *args, **kwargs: self(*args, **kwargs))
File "E:\stable-diffusion-webui\modules\sd_hijack_utils.py", line 28, in __call__
return self.__orig_func(*args, **kwargs)
File "E:\stable-diffusion-webui\repositories\stable-diffusion-stability-ai\ldm\models\diffusion\ddpm.py", line 858, in apply_model
x_recon = self.model(x_noisy, t, **cond)
File "E:\stable-diffusion-webui\venv\lib\site-packages\torch\nn\modules\module.py", line 1538, in _call_impl
result = forward_call(*args, **kwargs)
File "E:\stable-diffusion-webui\repositories\stable-diffusion-stability-ai\ldm\models\diffusion\ddpm.py", line 1335, in forward
out = self.diffusion_model(x, t, context=cc)
File "E:\stable-diffusion-webui\venv\lib\site-packages\torch\nn\modules\module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "E:\stable-diffusion-webui\extensions\sd-webui-fabric\scripts\patching.py", line 199, in new_forward
out = self._fabric_old_forward(x, timesteps, context, **kwargs)
File "E:\stable-diffusion-webui\repositories\stable-diffusion-stability-ai\ldm\modules\diffusionmodules\openaimodel.py", line 797, in forward
h = module(h, emb, context)
File "E:\stable-diffusion-webui\venv\lib\site-packages\torch\nn\modules\module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "E:\stable-diffusion-webui\repositories\stable-diffusion-stability-ai\ldm\modules\diffusionmodules\openaimodel.py", line 84, in forward
x = layer(x, context)
File "E:\stable-diffusion-webui\venv\lib\site-packages\torch\nn\modules\module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "E:\stable-diffusion-webui\repositories\stable-diffusion-stability-ai\ldm\modules\attention.py", line 334, in forward
x = block(x, context=context[i])
File "E:\stable-diffusion-webui\venv\lib\site-packages\torch\nn\modules\module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "E:\stable-diffusion-webui\repositories\stable-diffusion-stability-ai\ldm\modules\attention.py", line 269, in forward
return checkpoint(self._forward, (x, context), self.parameters(), self.checkpoint)
File "E:\stable-diffusion-webui\repositories\stable-diffusion-stability-ai\ldm\modules\diffusionmodules\util.py", line 121, in checkpoint
return CheckpointFunction.apply(func, len(inputs), *args)
File "E:\stable-diffusion-webui\venv\lib\site-packages\torch\autograd\function.py", line 506, in apply
return super().apply(*args, **kwargs) # type: ignore[misc]
File "E:\stable-diffusion-webui\repositories\stable-diffusion-stability-ai\ldm\modules\diffusionmodules\util.py", line 136, in forward
output_tensors = ctx.run_function(*ctx.input_tensors)
File "E:\stable-diffusion-webui\repositories\stable-diffusion-stability-ai\ldm\modules\attention.py", line 272, in _forward
x = self.attn1(self.norm1(x), context=context if self.disable_self_attn else None) + x
File "E:\stable-diffusion-webui\venv\lib\site-packages\torch\nn\modules\module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "E:\stable-diffusion-webui\extensions\sd-webui-fabric\scripts\patching.py", line 178, in patched_attn1_forward
out_cond = weighted_attention(attn1, attn1._fabric_old_forward, x_cond, ctx_cond, ws, **kwargs) # (n_cond, seq, dim)
File "E:\stable-diffusion-webui\extensions\sd-webui-fabric\scripts\weighted_attention.py", line 163, in weighted_attention
assert _xformers_attn in locals() or _xformers_attn in globals(), "xformers attention function not found"
AssertionError: xformers attention function not foundSecond run stacktrace:
Traceback (most recent call last):
File "E:\stable-diffusion-webui\modules\call_queue.py", line 57, in f
res = list(func(*args, **kwargs))
File "E:\stable-diffusion-webui\modules\call_queue.py", line 37, in f
res = func(*args, **kwargs)
File "E:\stable-diffusion-webui\modules\txt2img.py", line 57, in txt2img
processed = processing.process_images(p)
File "E:\stable-diffusion-webui\modules\processing.py", line 611, in process_images
res = process_images_inner(p)
File "E:\stable-diffusion-webui\extensions\sd-webui-controlnet\scripts\batch_hijack.py", line 42, in processing_process_images_hijack
return getattr(processing, '__controlnet_original_process_images_inner')(p, *args, **kwargs)
File "E:\stable-diffusion-webui\modules\processing.py", line 729, in process_images_inner
samples_ddim = p.sample(conditioning=p.c, unconditional_conditioning=p.uc, seeds=p.seeds, subseeds=p.subseeds, subseed_strength=p.subseed_strength, prompts=p.prompts)
File "E:\stable-diffusion-webui\extensions\sd-webui-fabric\scripts\marking.py", line 28, in process_sample
return process.sample_before_CN_hack(*args, **kwargs)
File "E:\stable-diffusion-webui\modules\processing.py", line 977, in sample
samples = self.sampler.sample(self, x, conditioning, unconditional_conditioning, image_conditioning=self.txt2img_image_conditioning(x))
File "E:\stable-diffusion-webui\modules\sd_samplers_kdiffusion.py", line 383, in sample
samples = self.launch_sampling(steps, lambda: self.func(self.model_wrap_cfg, x, extra_args={
File "E:\stable-diffusion-webui\modules\sd_samplers_kdiffusion.py", line 257, in launch_sampling
return func()
File "E:\stable-diffusion-webui\modules\sd_samplers_kdiffusion.py", line 383, in <lambda>
samples = self.launch_sampling(steps, lambda: self.func(self.model_wrap_cfg, x, extra_args={
File "E:\stable-diffusion-webui\venv\lib\site-packages\torch\utils\_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
File "E:\stable-diffusion-webui\repositories\k-diffusion\k_diffusion\sampling.py", line 626, in sample_dpmpp_2m_sde
denoised = model(x, sigmas[i] * s_in, **extra_args)
File "E:\stable-diffusion-webui\venv\lib\site-packages\torch\nn\modules\module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "E:\stable-diffusion-webui\modules\sd_samplers_kdiffusion.py", line 143, in forward
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]))
File "E:\stable-diffusion-webui\venv\lib\site-packages\torch\nn\modules\module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "E:\stable-diffusion-webui\repositories\k-diffusion\k_diffusion\external.py", line 112, in forward
eps = self.get_eps(input * c_in, self.sigma_to_t(sigma), **kwargs)
File "E:\stable-diffusion-webui\repositories\k-diffusion\k_diffusion\external.py", line 138, in get_eps
return self.inner_model.apply_model(*args, **kwargs)
File "E:\stable-diffusion-webui\modules\sd_hijack_utils.py", line 17, in <lambda>
setattr(resolved_obj, func_path[-1], lambda *args, **kwargs: self(*args, **kwargs))
File "E:\stable-diffusion-webui\modules\sd_hijack_utils.py", line 28, in __call__
return self.__orig_func(*args, **kwargs)
File "E:\stable-diffusion-webui\repositories\stable-diffusion-stability-ai\ldm\models\diffusion\ddpm.py", line 858, in apply_model
x_recon = self.model(x_noisy, t, **cond)
File "E:\stable-diffusion-webui\venv\lib\site-packages\torch\nn\modules\module.py", line 1538, in _call_impl
result = forward_call(*args, **kwargs)
File "E:\stable-diffusion-webui\repositories\stable-diffusion-stability-ai\ldm\models\diffusion\ddpm.py", line 1335, in forward
out = self.diffusion_model(x, t, context=cc)
File "E:\stable-diffusion-webui\venv\lib\site-packages\torch\nn\modules\module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "E:\stable-diffusion-webui\extensions\sd-webui-fabric\scripts\patching.py", line 156, in new_forward
_ = self._fabric_old_forward(zs, ts, ctx)
File "E:\stable-diffusion-webui\repositories\stable-diffusion-stability-ai\ldm\modules\diffusionmodules\openaimodel.py", line 797, in forward
h = module(h, emb, context)
File "E:\stable-diffusion-webui\venv\lib\site-packages\torch\nn\modules\module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "E:\stable-diffusion-webui\repositories\stable-diffusion-stability-ai\ldm\modules\diffusionmodules\openaimodel.py", line 84, in forward
x = layer(x, context)
File "E:\stable-diffusion-webui\venv\lib\site-packages\torch\nn\modules\module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "E:\stable-diffusion-webui\repositories\stable-diffusion-stability-ai\ldm\modules\attention.py", line 334, in forward
x = block(x, context=context[i])
File "E:\stable-diffusion-webui\venv\lib\site-packages\torch\nn\modules\module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "E:\stable-diffusion-webui\repositories\stable-diffusion-stability-ai\ldm\modules\attention.py", line 269, in forward
return checkpoint(self._forward, (x, context), self.parameters(), self.checkpoint)
File "E:\stable-diffusion-webui\repositories\stable-diffusion-stability-ai\ldm\modules\diffusionmodules\util.py", line 121, in checkpoint
return CheckpointFunction.apply(func, len(inputs), *args)
File "E:\stable-diffusion-webui\venv\lib\site-packages\torch\autograd\function.py", line 506, in apply
return super().apply(*args, **kwargs) # type: ignore[misc]
File "E:\stable-diffusion-webui\repositories\stable-diffusion-stability-ai\ldm\modules\diffusionmodules\util.py", line 136, in forward
output_tensors = ctx.run_function(*ctx.input_tensors)
File "E:\stable-diffusion-webui\repositories\stable-diffusion-stability-ai\ldm\modules\attention.py", line 272, in _forward
x = self.attn1(self.norm1(x), context=context if self.disable_self_attn else None) + x
File "E:\stable-diffusion-webui\venv\lib\site-packages\torch\nn\modules\module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "E:\stable-diffusion-webui\extensions\sd-webui-fabric\scripts\patching.py", line 140, in patched_attn1_forward
out = attn1._fabric_old_forward(x, **kwargs)
File "E:\stable-diffusion-webui\extensions\sd-webui-fabric\scripts\patching.py", line 178, in patched_attn1_forward
out_cond = weighted_attention(attn1, attn1._fabric_old_forward, x_cond, ctx_cond, ws, **kwargs) # (n_cond, seq, dim)
File "E:\stable-diffusion-webui\extensions\sd-webui-fabric\scripts\weighted_attention.py", line 154, in weighted_attention
if is_the_same(attn_fn, split_cross_attention_forward_invokeAI):
File "E:\stable-diffusion-webui\extensions\sd-webui-fabric\scripts\weighted_attention.py", line 147, in is_the_same
return fn1.__name__ == fn2.__name__ and fn1.__module__ == fn2.__module__
AttributeError: 'functools.partial' object has no attribute '__name__'I've been struggling getting xformers to work in the past, but I believe that this is an error I haven't seen before, I'll look into.
I also had a look at how the Tiled Diffusion plugin does the hijacking of the attention methods, including xformers. Having used it for months in thousands of image generations without any problems, I consider this plugin to be very robust. Please have a look at the following. It's an interesting alternative hijacking approach that may provide some insights and help solve other issues as well:
dvruette
commented
1 year ago Ok, I've pretty much overhauled the entire weighted attention code using a similar approach to the code you linked, which should hopefully make it more robust. Unfortunately, I still didn't have any luck reproducing your error, but I'm more confident now that it should be fixed now.
Lembont
commented
1 year ago I had the same error but updating the script seems to have fixed it ! Thank you so much !
h3rmit-git
commented
1 year ago It works!
Unlike with --opt-split-attention, which can only do 256x256 images with 4GB of VRAM, I can now generate 512x512 images with --xformers and 4 Liked images!
However, there's still one one more bug with the patching/unpatching process. In case of an error during the generation (e.g. out of memory), the next generation enters an infinite loop (100% CPU, 0% GPU, no progress).
dvruette
commented
1 year ago Awesome! I think I know what could cause the issue you’re describing, although I thought it was being handled correctly. Should be easier to reproduce as well, will have a look.
h3rmit-git
commented
1 year ago Thanks!
On the same topic of patching, there's also always a message about Fabric restoring the original U-Net, even when it's disabled. This should be relevant to the bug.
100%|██████████████████████████████████████████| 20/20 [00:56<00:00, 2.80s/it]
[FABRIC] Restoring original U-Net forward pass█| 20/20 [00:27<00:00, 1.46s/it]
Total progress: 100%|██████████████████████████| 20/20 [00:35<00:00, 1.77s/it]
100%|██████████████████████████████████████████| 20/20 [00:31<00:00, 1.60s/it]
[FABRIC] Restoring original U-Net forward pass█| 20/20 [00:27<00:00, 1.46s/it]
Total progress: 100%|██████████████████████████| 20/20 [00:33<00:00, 1.67s/it]
100%|██████████████████████████████████████████| 20/20 [00:32<00:00, 1.60s/it]
[FABRIC] Restoring original U-Net forward pass█| 20/20 [00:27<00:00, 1.47s/it]
Total progress: 100%|██████████████████████████| 20/20 [00:33<00:00, 1.66s/it]
100%|██████████████████████████████████████████| 20/20 [00:35<00:00, 1.77s/it]
[FABRIC] Restoring original U-Net forward pass█| 20/20 [00:31<00:00, 1.45s/it]
Total progress: 100%|██████████████████████████| 20/20 [00:36<00:00, 1.84s/it]
Total progress: 100%|██████████████████████████| 20/20 [00:36<00:00, 1.45s/it]```I managed to fix the infinite loop issue and opened a Pull Request: https://github.com/dvruette/sd-webui-fabric/pull/28
Edit: I also added a fix about the above U-Net restoration message in the same Pull Request.
I am getting the following error when running FABRIC with --xformers.
The error goes away with --opt-split-attention instead of --xformers.
Versions: