Added normal and latent inferers for ControlNet. Added tests (copied … #439
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…from the normal inferer tests, but with the addition of controlnet support).
ericspod
reviewed
Nov 28, 2023
generative/inferers/inferer.py
Outdated
Comment on lines
686
to
697
| if isinstance(diffusion_model, SPADEDiffusionModelUNet): | ||
| model_output = diffusion_model( | ||
| model_input, timesteps=torch.Tensor((t,)).to(input_noise.device), context=None, seg=seg, | ||
| down_block_additional_residuals=down_block_res_samples, | ||
| mid_block_additional_residual=mid_block_res_sample | ||
| ) | ||
| else: | ||
| model_output = diffusion_model( | ||
| model_input, timesteps=torch.Tensor((t,)).to(input_noise.device), context=None, | ||
| down_block_additional_residuals=down_block_res_samples, | ||
| mid_block_additional_residual=mid_block_res_sample | ||
| ) |
There was a problem hiding this comment.
Suggested change
| if isinstance(diffusion_model, SPADEDiffusionModelUNet): | |
| model_output = diffusion_model( | |
| model_input, timesteps=torch.Tensor((t,)).to(input_noise.device), context=None, seg=seg, | |
| down_block_additional_residuals=down_block_res_samples, | |
| mid_block_additional_residual=mid_block_res_sample | |
| ) | |
| else: | |
| model_output = diffusion_model( | |
| model_input, timesteps=torch.Tensor((t,)).to(input_noise.device), context=None, | |
| down_block_additional_residuals=down_block_res_samples, | |
| mid_block_additional_residual=mid_block_res_sample | |
| ) | |
| _seg = seg if isinstance(diffusion_model, SPADEDiffusionModelUNet) else None | |
| model_output = diffusion_model( | |
| model_input, timesteps=torch.Tensor((t,)).to(input_noise.device), context=None, seg=_seg, | |
| down_block_additional_residuals=down_block_res_samples, | |
| mid_block_additional_residual=mid_block_res_sample | |
| ) |
We could reduce the amount of code duplication with this sort of pattern, here and elsewhere.
There was a problem hiding this comment.
that exact formulation won't work as DiffusionModelUnet doesn't have the argument seg. But this would work:
diffusion_model = partial(diffusion_model, seg=seg) if isinstance(diffusion_model, SPADEDiffusionModelUNet) else diffusion_model
model_output = diffusion_model(
model_input, timesteps=torch.Tensor((t,)).to(input_noise.device), context=None,
down_block_additional_residuals=down_block_res_samples,
mid_block_additional_residual=mid_block_res_sample(would need from functools import partial up top)
marksgraham
requested changes
Nov 30, 2023
generative/inferers/inferer.py
Outdated
| """ | ||
| ControlNetDiffusionInferer takes a trained diffusion model and a scheduler and can be used to perform a signal | ||
| forward pass for a training iteration, and sample from the model, supporting ControlNet-based conditioning. | ||
| Args: |
There was a problem hiding this comment.
please add a line between the intro and Args
tests/test_controlnet_inferers.py
Outdated
| ], | ||
| ] | ||
|
|
||
| class CN_TestDiffusionSamplingInferer(unittest.TestCase): |
There was a problem hiding this comment.
This needs to use CapsWord convention, i.e. remove the underscore. This gets picked up when you run formatting tests with ./runtests.sh --codeformat --autofix
tests/test_controlnet_inferers.py
Outdated
| ) | ||
| self.assertEqual(len(intermediates), 10) | ||
|
|
||
| class LCN_TestDiffusionSamplingInferer(unittest.TestCase): |
There was a problem hiding this comment.
CapsWords convention also required here
generative/inferers/inferer.py
Outdated
Comment on lines
686
to
697
| if isinstance(diffusion_model, SPADEDiffusionModelUNet): | ||
| model_output = diffusion_model( | ||
| model_input, timesteps=torch.Tensor((t,)).to(input_noise.device), context=None, seg=seg, | ||
| down_block_additional_residuals=down_block_res_samples, | ||
| mid_block_additional_residual=mid_block_res_sample | ||
| ) | ||
| else: | ||
| model_output = diffusion_model( | ||
| model_input, timesteps=torch.Tensor((t,)).to(input_noise.device), context=None, | ||
| down_block_additional_residuals=down_block_res_samples, | ||
| mid_block_additional_residual=mid_block_res_sample | ||
| ) |
There was a problem hiding this comment.
that exact formulation won't work as DiffusionModelUnet doesn't have the argument seg. But this would work:
diffusion_model = partial(diffusion_model, seg=seg) if isinstance(diffusion_model, SPADEDiffusionModelUNet) else diffusion_model
model_output = diffusion_model(
model_input, timesteps=torch.Tensor((t,)).to(input_noise.device), context=None,
down_block_additional_residuals=down_block_res_samples,
mid_block_additional_residual=mid_block_res_sample(would need from functools import partial up top)
generative/inferers/inferer.py
Outdated
| else: | ||
| return total_kl | ||
|
|
||
| def _approx_standard_normal_cdf(self, x): |
There was a problem hiding this comment.
if we inherit from DiffusionModelInferer we can remove this function here
generative/inferers/inferer.py
Outdated
| torch.sqrt(torch.Tensor([2.0 / math.pi]).to(x.device)) * (x + 0.044715 * torch.pow(x, 3))) | ||
| ) | ||
|
|
||
| def _get_decoder_log_likelihood( |
There was a problem hiding this comment.
this can also be removed if we inherit from DiffusionModelInfererer
added 4 commits
November 30, 2023 15:24
…E check. Fixed some formatting typos. Deleted two functions and changed inheritance of ControlNetDiffusionInferer. Changed names of tests to agree with caps convention.
…E check. Fixed some formatting typos. Deleted two functions and changed inheritance of ControlNetDiffusionInferer. Changed names of tests to agree with caps convention. + run autofix
…ials that had been forgotten last commit.
marksgraham
approved these changes
Dec 1, 2023
marksgraham
deleted the
438-create-a-new-latentdiffusioninferer-compatible-with-controlnet
branch
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
…from the normal inferer tests, but with the addition of controlnet support).