Sure i can try to help with LoRA generation, even that i am still a beginner and dont know what works best.

For this LoRA
I used the KohyaSS Kolab:
https://colab.research.google.com/github/Linaqruf/kohya-trainer/blob/main/kohya-LoRA-dreambooth.ipynb#scrollTo=Po6HdriyM7oC

At 1.1 and 1.2 and 1.3 i didnt change anything and just hit the run button

And 1.3.1 is just putting your huggingface token for model download.

1.3.2 is for gdrive connections.

I did run 1.4 but i dont know what the file explorer does

**At 2.1 i selected the anything 4.5 for training**

I didnt run 2.2

2.3 needs to be run with the anime.vae.pt

3.1 i didnt change anything leave it at 1 and style

** At 3.2 i changed train count from 5 to 1, you can leave mksks and the style

3.3 you dont need to change anything

4.1 i didnt run this one

4.2 i dint run this one

4.3 is the tag to the character, you want to train.
I left it at gelbooru and added
lyn_(fire_emblem)
To the tag 1
Tag2 i left empty**

5.1 i didnt run it

5.2 i didnt run it

**6.1 run this and check the keep Metadata checkbox**

6.2 i didnt run this

** 7.1 i didnt check v2 or v_parametatzion
Projekt name choose what you want i choose LynFireEmblem
Change the pretrained model name to
content/pre_trained_model/Anything-v4-5-pruned.safetensors

Change the vae to the anime.vae.pt in your content folder i dont know the link anymore

Train folder, reg folder and output dir you can leave at it is.

Check the output to drive checkbox, it saves the models into your drive

7.2 leave everything as it is and run it

7.3 train batch size is 4
I changed the number of epochs from 1 to 15
Everything is the same except lr_shedluer change it to cosine with restarts
Run the cell it will take a lot of time about 3hours.

After its finished check your results in the webui at my tests itbwas mostly between 9 and 13 epochs the best results.
Hope this helps, if you have any questions there is a discord group called Waifu Diffusion with a lot of nice people that can help you.