Pinned

1. openai/evals openai/evals Public

   Evals is a framework for evaluating LLMs and LLM systems, and an open-source registry of benchmarks.

   Python 14.1k 2.5k

2. summarizing-from-human-feedback summarizing-from-human-feedback Public

   Implementation of OpenAI's "Learning to Summarize with Human Feedback"

   Jupyter Notebook 6

3. bitblaster-16 bitblaster-16 Public

   BitBlaster-16 is a 16-bit computer built from scratch using only NAND gates and data flip-flops as primitives! :)

   Python 1

4. fermi-poker fermi-poker Public

   Learn how to quantify your uncertainty by playing Fermi poker!

   Python 2

5. self-taught-critiquer self-taught-critiquer Public

   Reducing the time to create critique-writing models by 100-1000x on n-digit arithmetic problems by getting the model to learn from its own generated outputs.

   Python 1

6. A PyTorch hook that's registered in... A PyTorch hook that's registered in a `with` statement

   1

   # This hook is particularly useful when ablating layers

   2

   3

   class ContextHook:

   4

       def __init__(self, layer):

   5

           self.layer = layer