Wagner Bruna

I've reviewed and tested a few of the smaller PRs that had no reviews / approvals. These look good to me: #316, #422, #468 , #591, #629 . These IMHO...

Tried to reproduce it here (gfx1102 (RX 7600 XT) on Linux), and it's even worse: black image from start with any Qwen variant (Pruned, Edit, etc). Same command line works...

No failures with my card: all tests either pass, or are not supported.

> Tried to reproduce it here (gfx1102 (RX 7600 XT) on Linux), and it's even worse: black image from start with any Qwen variant (Pruned, Edit, etc). I was able...

APG works even with distilled models. I was able to get good LCM generations with 4+ CFG, and _negatives_.

> I think the naming doesn’t really matter. Once the potential license issue I mentioned in the review comments is resolved, this PR can be merged. @leejet , your comments...

It could be argued that the matrices are just the product of an algorithm (training, a simple least-squares approximation, etc), and thus not restricted by copyright. The problem is the...

This has good intentions, but the PR has many unrelated and inappropriate changes, like this one: [](https://github.com/leejet/stable-diffusion.cpp/pull/457/commits/f7717260d3146cd1136f994193b8c6d659965706#diff-1d95451d72f080678e08278a6d33203dbc4d9502f688ab507f9ac806c8a78bddL738) And things that look a lot like something an LLM would mess up...

I'd suggest testing the patch, and posting about the results on the PR.

> "prompt": "A lovely cat{\"seed\": 357925}", If the idea is to be able to type it into the prompt of an existing application, maybe it could be simplified a bit?...