Better image interpolation in training scripts follow up

[asomoza]

With #11206 we did a small quality improvement for the SDXL Dreambooth LoRA script by making LANCZOS the default interpolation mode for the image resizing.

This issue is to ask for help from the community to bring this change to the other training scripts, specially for the popular ones.

Since this is a really easy to make contribution I'll ask that we leave this issue for beginners and people that want to start learning how to contribute to open source projects.

What I think are the most important ones:

• train_dreambooth_flux
• train_dreambooth_lora
• train_dreambooth_lora_lumina2
• train_dreambooth_lora_sdxl
• train_controlnet_flux
• train_controlnet_sdxl
• train_text_to_image
• train_text_to_image_lora
• train_text_to_image_sdxl
• train_text_to_image_lora_sdxl
• train_dreambooth_lora_flux_advanced
• train_dreambooth_lora_sd15_advanced
• train_dreambooth_lora_sdxl_advanced

If you have other preference, please feel free to ask me to add it.

If you want to contribute just answer to this issue with the one you want to do and tag me in the PR. Please only take one since I want to use this issue to get people to learn the ropes on how to contribute and get started with open source.

[ishandutta0098]

Hi @asomoza I would like to contribute to train_dreambooth_flux

[merterbak]

Hi, I want to contribute to train_dreambooth_lora

[tongyu0924]

Hi, I want to contribute to train_text_to_image

[YoulunPeng]

Hi, all! I will help with this one train_controlnet_flux

Here is the PR #11449