Vision Model Pre-training on Interleaved Image-Text Data via Latent Compression Learning - View it on GitHub