Conversation
atiorh
left a comment
atiorh
left a comment
There was a problem hiding this comment.
Thanks for submitting this PR @hanstjua! I left some early comments and requested some changes. Before fully reviewing the math, I tested generating an image with the CLI and was only able to generate noise. Could you please address the comments I left and verify that you are able to generate an image for the prompt "a high quality photo of a surfing dog" as a sanity check before a second pass? Thanks again!
|
|
||
| self.initNoiseSigma = sigmas.max()! | ||
|
|
||
| self.timeSteps = linspace(0, Float(trainStepCount - 1), trainStepCount).reversed().map { Int(round($0)) } |
There was a problem hiding this comment.
Shouldn't the length of self.timeSteps adhere to self.inferenceStepCount instead of trainStepCount?
There was a problem hiding this comment.
Yes, you are right. I'll make this change.
| switch config.schedulerType { | ||
| case .pndmScheduler: return PNDMScheduler(stepCount: config.stepCount) | ||
| case .dpmSolverMultistepScheduler: return DPMSolverMultistepScheduler(stepCount: config.stepCount) | ||
| case .eulerAncestralDiscreteScheduler: return EulerAncestralDiscreteScheduler(randomSource: randomSource(from: config.rngType, seed: config.seed)) |
There was a problem hiding this comment.
The new scheduler does not take config.stepCount as an argument, is this expected?
There was a problem hiding this comment.
Related: Could you please register this scheduler in the CLI enum as well?
There was a problem hiding this comment.
No, this is not expected. Seems like I missed this. Thanks for pointing out.
I'll register it in CLI enum.
There was a problem hiding this comment.
Great, looking forward to it :)
There was a problem hiding this comment.
I've made the changes, but still couldn't seem to generate a proper image. Currently still looking into what's the issue.
2 participants
Thank you for your interest in contributing to Core ML Stable Diffusion! Please review CONTRIBUTING.md first. If you would like to proceed with making a pull request, please indicate your agreement to the terms outlined in CONTRIBUTING.md by checking the box below. If not, please go ahead and fork this repo and make your updates.
We appreciate your interest in the project!
Do not erase the below when submitting your pull request:
#########