bilinear instead of bicubic

add safetensors from google model

fix issues with erf and xavier init

fix config

use transformers siglip modeling implementation except for flash attention

add convert siglip to hf

add processing/tokenization siglip

fix configuration imports

make processing similar to transformers implementation

make config similar to transformers except for flash

Update config.json

Upload folder using huggingface_hub

initial commit