What happened to the chroma rapid
Sorry for the wrong thread, but I loved that project.
I loved it too. Great model. I would love to still share and update it.
The author of the 2k-test LORA I was using from an Apache 2.0-labeled repository updated their model card license at the same time they reported I was using it incorrectly (see https://huggingface.co/silveroxides/Chroma-LoRAs/commit/c0eff3e0b7eedde205cd7bcd87eebf41154a644c). This was likely done to reflect the "development only" restrictive license repo where is was sourced from: https://huggingface.co/lodestones/chroma-debug-development-only (although it says it will be uploaded to a separate repo with an Apache 2.0 license when ready, hence me getting it from the Apache 2.0 separate repo).
Anyway, I was happy to give credit and make modifications, but the 2k-test author just wanted me to take it down and believed I wasn't doing it right. Other related Chroma resources I was hoping to use have been deleted from Huggingface. In order to respect the desires of the author, I took it down.
Overall it was perhaps a miscommunication and difference of opinion on how models should be packaged and distributed and I hold no ill will towards the Chroma community. I'm planning on moving on to Qwen Image which ironically seems more open. I wish the Chroma community the best of luck, as fundamentally, it is a pretty good model.
EDIT: Closing this thread, as I want this community discussion to be related to the WAN AIO. Feel free to join Discord if you want to talk further.
What a shame, I was about to make a reddit post leading newcomers to this model as the best solution for chroma. I had great results with (euler_ancestral, ddim_uniform, 2CFG, 12 steps )
Thank you for the detailed explanation and your contributions.
I have to say, seeing some of Silveroxide's posts and responses lately to legitimate questions, I can believe this entirely...
Unfortunately, since trying to use their models with their new custom loader node, all I get from Chroma is 99% VRAM usage and awful outputs. Does make me wonder who it is who is 'not using it correctly'...
Chroma has been my GOAT for months now, since around Epoch 30. But sadly and ironically, since v50 / Chroma1HD, it's all gone to crap... Excessive VRAM usage, impossible to generate anything without some form of accelerator that then impacts prompt adherence and output style... and don't even get me started on Radiance... Unless you have a 7000GB VRAM GPU, trying to run that is like deliberately taunting your PC.