This is a Flux.1 dev full finetune modeluse nearly 30000 image in 120+a100 hours tunedThis is still an early attempt for me to make large-scale adjustments to the FLUX modelObviously, it seems that we have achieved quite good results this timePlease note that when using the model, please use CFG parameters and add negative promptsFine tuning seems to have reconstructed some CFGPlease control CFG at 3.5 for optimal resultsThe model has been integrated into FP8's checkpoint, and can be directly loaded using the load checkpoint without the need to separately load clips and t5xxlI think these parameters are distilled parameters of the model. Should they be retained or removed in large-scale trainingI think it needs to be removed. Experienced friends are welcome to reply and discussSo, let's go back in timeenjoy it!

Files with SHA256: 505683a8e7ee8640e108fc869f59bd607a2757c96e6f67200e34c61c73134324