Original Summary

Imagine Git for model fine-tunes that also saves you storage.<p>DeltaTensors compresses fine-tuned model checkpoints by storing the weight difference from the base model instead of another full copy of the model.<p>If you have a bunch of fine-tunes or checkpoints from the same base model, you keep the base once and store a much smaller .wdelta for each version.<p>On a Qwen2.5-0.5B fine-tune, a 953 MB model compressed to a 294 MB delta. Reconstructing it changed perplexity from 19.11 to 19.22.<p>It works after training, so you do not need to use LoRA or change how you fine-tune the model. You can take an existing full fine-tune and compress it afterward.<p>It also supports chained deltas for model version history, Hugging Face Trainer checkpoints, and streaming compression&#x2F;reconstruction so you do not need to load two full models into memory.<p>https:&#x2F;&#x2F;github.com&#x2F;AaravGaurdev&#x2F;deltatensors

  • 情报分类：技术学习与提效
  • 分类依据：内容涉及技术、AI、软件工具或工程实践
  • 信息来源：Hacker News 新项目
  • 发布时间：2026/9/21 13:25:32