You can test the weights of FLUX LoRA. This is a cyclic workflow where each of the 0-58 blocks is set to 0 individually. Through this process, you can test the impact of each block on the final result.

This is an experimental test, the results are based on the Inspire pack. However, in my tests, different blocks do have varying impacts on the results, though the test outcomes lack stability.

Since Flux is based on the DIT framework, the definition of layers is also quite vague, making it difficult to obtain a precise result. It's also hard to summarize any regular patterns from different fixed-concept LoRAs, so testing might be in vain. However, there may still be some patterns to discover within a single LoRA concept.

你可以测试flux lora的权重

这是一个循环的工作流程，分别让0-58个块单独为0。通过这套流程，你可以测试不同块对最终结果的影响。

这是一个具有实验性的测试，这是根据inspire pack得到的结果。但在我的测试下不同块确实会对结果产生不同的影响，虽然测试结果不具备稳定性。

这是Flux_LoRA的61个块，但最终前三个块被排除后只剩58个块

This is 61- 3 layers

CLIP  T5 IN D00 D01 D02 D03 D04 D05 D06 D07 D08 D09 D10 D11 D12 D13 D14 D15 D16 D17 D18 S00 S01 S02 S03 S04 S05 S06 S07 S08 S09 S10 S11 S12 S13 S14 S15 S16 S17 S18 S19 S20 S21 S22 S23 S24 S25 S26 S27 S28 S29 S30 S31 S32 S33 S34 S35 S36 S37 OUT

并且flux基于DIT框架，所以对层的定义也比较模糊，这样大家就很难得到一个非常确切的结果，也很难根据不同固定概念的lora总结出一些规律性经验，所以测试可能是徒劳的，但单个lora的概念或许还是有些规律可以找到

需要单独下载节点，目前还未同步更新至Manager

https://github.com/SS-snap/ComfyUI-LBW_flux.git