Jipeng Zhang
Jipeng Zhang
Add support for Galactica model would be very helpful. It seems like the most powerful full open-source LLM in MMLU benchmark.
I tried to extract output prediction probs from pytorch_backend.py in flex_opt. But I can not get logits returns. I have little idea about extracting logits from your complex distribution computations....
Following the script provided in the second step of "Selecting data for a task" in your readme, I have a command line that needs to be run as shown below:...