Closed Vadimkomis closed 4 months ago
Hi @Vadimkomis. Sorry for any confusion, but Dreambooth doesn't use the training API. See this blog post for details: https://replicate.com/blog/dreambooth-api
As that blog post mentions, SDXL fine-tuning is a better way to achieve this, and that uses the training API. So I'd recommend using that.
@mattt thanks for the quick response. I use the replicate-swift sdk, and that returns the regular api of https://api.replicate.com/v1/models/replicate/dreambooth/versions/cd3f925f7ab21afaef7d45224790eedbb837eeac40d22e8fefe015489ab644aa/trainings
and not the experimental endpoint of https://dreambooth-api-experimental.replicate.com/v1/trainings
. Is there a way to get that endpoint from replicate-swift?
@Vadimkomis No, the Swift library doesn't support the experimental Dreambooth API. Again, I strongly recommend SDXL fine-tuning, which is supported by createTraining
and produces much better results.
I run the following code from my app and I keep getting ""Internal server error""
_ = try await replicate.createTraining(model: StableDiffusion.modelID, version: StableDiffusion.versionID, destination: StableDiffusion.destination, input: input)
This is the full URL:
https://api.replicate.com/v1/models/replicate/dreambooth/versions/cd3f925f7ab21afaef7d45224790eedbb837eeac40d22e8fefe015489ab644aa/trainings
Here are the params: ` Optional<Dictionary<String, Value>> ▿ some : 2 elements ▿ 0 : 2 elements