A question about "network_weight"

[yli6hz]

For flux lora, sometimes I find my lora training result seems not good at all, then I try to add some more steps base on finished lora file. For Kohya Scripts, I just input the file path and file name such as "E:\fluxgym\output\XXX.safetensors" to fill in "network_weight" and start up again via parameters modification. But I find the result shows a damage of whole person. It doesn't work.

What is correct way to use previous trained file in fluxgym?
Same question about "base_weight" which is to be used in differentiation to speed up lora training. Does fluxgym support this?

Another question is about optimizer. I find there are many optimizers such as "prodigy, SGDNesterov" in others training script, but there are three kinds in fluxgym. Is it possible to add more?

Thanks!