Add moge_v2_int8.tflite (post-training quantized from moge_v2.tflite)

Add moge_v2_fp16.tflite (post-training quantized from moge_v2.tflite)

Upload moge_v2.tflite with huggingface_hub

Keep only weights in HF repo; source/config/README live in GitHub (1kaiser/astro)

Delete README.md

revert: remove notebook from HF repo

use uv instead of pip for deps

add model card

add config

upload weights (attempt 3)

add reconstruct.py CLI/lib

add inference source files

initial commit