Florence2

๐Ÿ”ฅFlorence-2 Fine-tuning: Microsoft's foundation VLM just got more awesome! ๐Ÿš€ Learn about the fine-tuning and access useful links๐Ÿ‘‡

  • Team successfully fine-tuned Florence2 for DocVQA

  • 0 to 57.0 similarity score on validation set ๐Ÿ“ˆ

  • Works with limited resources (single GPU, small batch size)

  • Frozen vision encoder still yields great results

  • Fine-tuning code & Colab notebook available ๐Ÿค—

  • Demo is up on Hugging Face Spaces

Who's ready to fine-tune Florence-2 for their own tasks? ๐Ÿ”ฅ

๐Ÿ™Œ Massive props to team for making this accessible! @mervenoyann @andi_marafioti @skalskip92 Blog: https://lnkd.in/dVE2pQJb Colab: https://lnkd.in/dEnhFgiX Demo: https://lnkd.in/dQvqGA4g

Last updated