PotatoSpudowski / fastLLaMa

fastLLaMa: An experimental high-performance framework for running Decoder-only LLMs with 4-bit quantization in Python using a C/C++ backend.

Home Page:https://potatospudowski.github.io/fastLLaMa/

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Fix multiple relative pointer transform

amitsingh19975 opened this issue · comments

We don't track if we applied the relative pointer transform on a pointer when we save or load. Since multiple variables can be aliased to the same pointer, we may apply the transform more than once.

Good catch!

Should look into this when we get bandwidth!