History

Awni Hannun 88d7b67e6e add cache + generation, clean up some stuff		2023-12-13 22:26:33 -08:00
..
__init__.py	phi-2 draft	2023-12-13 22:23:38 -05:00
.gitignore	add cache + generation, clean up some stuff	2023-12-13 22:26:33 -08:00
convert.py	add cache + generation, clean up some stuff	2023-12-13 22:26:33 -08:00
hf_model.py	phi-2 draft	2023-12-13 22:23:38 -05:00
model.py	add cache + generation, clean up some stuff	2023-12-13 22:26:33 -08:00
phi2_outputs.txt	phi-2 draft	2023-12-13 22:23:38 -05:00
README.md	phi-2 draft	2023-12-13 22:23:38 -05:00
requirements.txt	add cache + generation, clean up some stuff	2023-12-13 22:26:33 -08:00

README.md

Phi-2

Phi-2 is a 2.7B parameter model released by Microsoft and trained on a mixture of GPT-4 outputs and clean web-text. Its performance theoretically rivals much, much stronger models.

Downloading and Converting Weights

To download and convert the model:

python phi2/convert.py

That will fill in weights/phi-2.npz.

Running the Model

🚧 (Not yet done) To run the model:

python phi2/generate.py

Layer-by-layer forward pass outputs are currently shown in the outputs.txt files.