Horace He
@cHHillee
가입 February 2010
633 팔로잉 중 55.4K 팬
Your model behaves differently during training and inference? Boy do I have a solution for you :^)
Training the model to detect when it is in training vs. deployment works too! The model can even generalize from vLLM deployment training to SGLang deployment. We do this by recording activations from layer 34 in Hugging Face Trainer's training step and in vLLM's compiled serving. Then, we swap in the attention and residual stream from the Hugging Face copy and SFT the model to predict "Yes," and to predict "No" when we swap in from vLLM instead. The model learns how its training vs. deployment environment affects its first 34 layers.
더 보기
