This repository was archived by the owner on Jun 22, 2026. It is now read-only.
fix(loader): cast model to fp16 before moving it to the device#56
Merged
Conversation
load_diffusion_cond moved the full fp32 model onto the GPU and then halved it to fp16. Casting to fp16 on the CPU first means only the fp16 model is transferred to the device. Measured A/B on the medium model, same seed through both orderings: output is bit-identical (max|new-old| = 0.0 across two runs); the torch-reported peak load allocation is 4.65 GB with the new order vs 9.35 GB with the old. fp32->fp16 rounds identically on CPU or GPU, so generation is unchanged.
This file contains hidden or 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
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
load_diffusion_cond moved the full fp32 model onto the GPU and then halved it to fp16. Casting to fp16 on the CPU first means only the fp16 model is transferred to the device.
Measured A/B on the medium model, same seed through both orderings: output is bit-identical (max|new-old| = 0.0 across two runs); the torch-reported peak load allocation is 4.65 GB with the new order vs 9.35 GB with the old. fp32->fp16 rounds identically on CPU or GPU, so generation seems unchanged.