mirror of
https://github.com/wassname/GENIES.git
synced 2026-06-27 16:10:25 +08:00
added data downloading
This commit is contained in:
@@ -2,3 +2,5 @@ models/
|
||||
configs/credentials.json
|
||||
build/
|
||||
env/
|
||||
upload_data.py
|
||||
genies-datasets.tar
|
||||
|
||||
@@ -8,7 +8,7 @@ As AI systems become more capable and are deployed in complex environments, it m
|
||||
## Quickstart
|
||||
|
||||
This repository contains:
|
||||
- Our datasets (`./distributions`) along with pairing specifications (`./distribution_shifts`).
|
||||
- Our datasets (`./distributions`) along with pairing specifications (`./distribution_shifts`). Download our datasets [here](https://genies-data.s3.us-east-2.amazonaws.com/genies-datasets.tar) or run the setup command after cloning the repo.
|
||||
- Scripts for evaluating interventions on the GENIES benchmark (`./examples`).
|
||||
- Our results (`./results`).
|
||||
- Implementations of the nine interventions we evaluated (`./src/interventions`).
|
||||
@@ -18,7 +18,7 @@ All of the models we fine-tuned with Lora can be found on [huggingface](https://
|
||||
**Setup:**
|
||||
```
|
||||
conda create --name env python=3.10
|
||||
pip install -e .
|
||||
pip install .
|
||||
python download_model_from_hf.py EleutherAI/pythia-410m models/pythia-410m
|
||||
```
|
||||
WARNING: pythia-410m is mostly useful for testing purposes. Most tuning interventions perform poorly with this model.
|
||||
|
||||
Binary file not shown.
@@ -0,0 +1,43 @@
|
||||
import requests
|
||||
import tarfile
|
||||
import os
|
||||
import fire
|
||||
|
||||
def download_tarfile(filename : str, download_dir : str):
|
||||
# URL of the file to download
|
||||
url = f"https://genies-data.s3.us-east-2.amazonaws.com/{filename}"
|
||||
|
||||
# Specify the local file path where you want to save the downloaded file
|
||||
downloaded_file_path = filename
|
||||
|
||||
# Specify the directory where you want to extract the contents
|
||||
extracted_dir = download_dir
|
||||
|
||||
# Download the file from the URL
|
||||
response = requests.get(url)
|
||||
|
||||
# Check if the request was successful (status code 200)
|
||||
if response.status_code == 200:
|
||||
# Save the downloaded content to a local file
|
||||
with open(downloaded_file_path, 'wb') as file:
|
||||
file.write(response.content)
|
||||
|
||||
print(f"File '{downloaded_file_path}' has been downloaded successfully.")
|
||||
|
||||
# Create the directory for extraction if it doesn't exist
|
||||
os.makedirs(extracted_dir, exist_ok=True)
|
||||
|
||||
# Extract the contents of the tar file
|
||||
with tarfile.open(downloaded_file_path, "r") as tar:
|
||||
tar.extractall(path=extracted_dir)
|
||||
|
||||
print(f"Contents of '{downloaded_file_path}' have been extracted to '{extracted_dir}'.")
|
||||
else:
|
||||
print(f"Failed to download the file. Status code: {response.status_code}")
|
||||
|
||||
def download_data():
|
||||
print("Downloading distributions...")
|
||||
download_tarfile("genies-datasets.tar", ".")
|
||||
|
||||
if __name__ == "__main__":
|
||||
fire.Fire(download_data)
|
||||
+120
-154
@@ -1,237 +1,203 @@
|
||||
absl-py==1.4.0
|
||||
accelerate==0.22.0
|
||||
aiohttp==3.8.5
|
||||
absl-py==2.0.0
|
||||
accelerate==0.24.1
|
||||
aiohttp==3.8.6
|
||||
aiosignal==1.3.1
|
||||
altair==5.1.2
|
||||
ansiwrap==0.8.4
|
||||
anyio==4.0.0
|
||||
annotated-types==0.6.0
|
||||
anyio==3.7.1
|
||||
appdirs==1.4.4
|
||||
argon2-cffi==23.1.0
|
||||
argon2-cffi-bindings==21.2.0
|
||||
arrow==1.2.3
|
||||
astor==0.8.1
|
||||
asttokens==2.4.0
|
||||
arrow==1.3.0
|
||||
asttokens==2.4.1
|
||||
async-lru==2.0.4
|
||||
async-timeout==4.0.3
|
||||
attrs==23.1.0
|
||||
awscli==1.29.43
|
||||
Babel==2.12.1
|
||||
backcall==0.2.0
|
||||
base58==2.1.1
|
||||
baukit @ git+https://github.com/davidbau/baukit@5e23007c02fd58f063200c5dc9033e90f092630d
|
||||
Babel==2.13.1
|
||||
bcrypt==4.0.1
|
||||
beautifulsoup4==4.12.2
|
||||
bitsandbytes==0.41.1
|
||||
black==21.12b0
|
||||
bleach==6.0.0
|
||||
blinker==1.6.3
|
||||
boto3==1.28.62
|
||||
botocore==1.31.62
|
||||
Brotli==1.1.0
|
||||
cachetools==5.3.2
|
||||
bitsandbytes==0.41.2.post2
|
||||
bleach==6.1.0
|
||||
boto3==1.28.84
|
||||
botocore==1.31.84
|
||||
certifi==2023.7.22
|
||||
cffi==1.15.1
|
||||
charset-normalizer==3.2.0
|
||||
click==7.1.2
|
||||
cmake==3.27.4.1
|
||||
colorama==0.4.4
|
||||
comm==0.1.4
|
||||
contourpy==1.1.0
|
||||
cryptography==41.0.3
|
||||
cycler==0.11.0
|
||||
cffi==1.16.0
|
||||
charset-normalizer==3.3.2
|
||||
click==8.1.7
|
||||
comm==0.2.0
|
||||
contourpy==1.2.0
|
||||
cryptography==41.0.5
|
||||
cycler==0.12.1
|
||||
Cython==0.29.36
|
||||
datasets==2.14.5
|
||||
debugpy==1.7.0
|
||||
datasets==2.14.6
|
||||
debugpy==1.8.0
|
||||
decorator==5.1.1
|
||||
deepspeed==0.10.2
|
||||
deepspeed==0.12.2
|
||||
defusedxml==0.7.1
|
||||
dill==0.3.7
|
||||
distro==1.8.0
|
||||
docker-pycreds==0.4.0
|
||||
docutils==0.16
|
||||
einops==0.6.1
|
||||
docopt==0.6.2
|
||||
docstring-parser==0.15
|
||||
einops==0.7.0
|
||||
entrypoints==0.4
|
||||
exceptiongroup==1.1.3
|
||||
executing==1.2.0
|
||||
fastjsonschema==2.18.0
|
||||
filelock==3.12.3
|
||||
executing==2.0.1
|
||||
fastjsonschema==2.18.1
|
||||
filelock==3.13.1
|
||||
fire==0.5.0
|
||||
flake8==6.1.0
|
||||
fonttools==4.42.1
|
||||
fonttools==4.44.0
|
||||
fqdn==1.5.1
|
||||
frozenlist==1.4.0
|
||||
fsspec==2023.6.0
|
||||
gitdb==4.0.10
|
||||
GitPython==3.1.35
|
||||
fsspec==2023.10.0
|
||||
gitdb==4.0.11
|
||||
GitPython==3.1.40
|
||||
h11==0.14.0
|
||||
hjson==3.1.0
|
||||
huggingface-hub==0.16.4
|
||||
httpcore==1.0.2
|
||||
httpx==0.25.1
|
||||
huggingface-hub==0.17.3
|
||||
idna==3.4
|
||||
inflate64==0.3.1
|
||||
iniconfig==2.0.0
|
||||
ipykernel==6.25.2
|
||||
ipython==8.15.0
|
||||
ipython-genutils==0.2.0
|
||||
ipywidgets==8.1.0
|
||||
ipykernel==6.26.0
|
||||
ipython==8.17.2
|
||||
ipywidgets==8.1.1
|
||||
isoduration==20.11.0
|
||||
isort==5.8.0
|
||||
jedi==0.19.0
|
||||
jedi==0.19.1
|
||||
Jinja2==3.1.2
|
||||
jmespath==1.0.1
|
||||
joblib==1.3.2
|
||||
json5==0.9.14
|
||||
jsonpointer==2.4
|
||||
jsonschema==4.19.0
|
||||
jsonschema==4.19.2
|
||||
jsonschema-specifications==2023.7.1
|
||||
jupyter==1.0.0
|
||||
jupyter-console==6.6.3
|
||||
jupyter-events==0.7.0
|
||||
jupyter-events==0.9.0
|
||||
jupyter-lsp==2.2.0
|
||||
jupyter_client==8.3.1
|
||||
jupyter_core==5.3.1
|
||||
jupyter_server==2.7.3
|
||||
jupyter_client==8.6.0
|
||||
jupyter_core==5.5.0
|
||||
jupyter_server==2.10.0
|
||||
jupyter_server_terminals==0.4.4
|
||||
jupyterlab==4.0.5
|
||||
jupyterlab==4.0.8
|
||||
jupyterlab-pygments==0.2.2
|
||||
jupyterlab-widgets==3.0.8
|
||||
jupyterlab_server==2.24.0
|
||||
jupyterlab-widgets==3.0.9
|
||||
jupyterlab_server==2.25.1
|
||||
kiwisolver==1.4.5
|
||||
lit==16.0.6
|
||||
markdown-it-py==3.0.0
|
||||
MarkupSafe==2.1.3
|
||||
matplotlib==3.7.2
|
||||
matplotlib==3.8.1
|
||||
matplotlib-inline==0.1.6
|
||||
mccabe==0.7.0
|
||||
mistune==3.0.1
|
||||
mpi4py @ file:///croot/mpi4py_1671223370575/work
|
||||
mdurl==0.1.2
|
||||
mistune==3.0.2
|
||||
mpmath==1.3.0
|
||||
multidict==6.0.4
|
||||
multiprocess==0.70.15
|
||||
multivolumefile==0.2.3
|
||||
mypy-extensions==1.0.0
|
||||
nbclient==0.8.0
|
||||
nbconvert==7.8.0
|
||||
nbclient==0.9.0
|
||||
nbconvert==7.11.0
|
||||
nbformat==5.9.2
|
||||
nest-asyncio==1.5.7
|
||||
networkx==3.1
|
||||
ninja==1.11.1
|
||||
nest-asyncio==1.5.8
|
||||
networkx==3.2.1
|
||||
ninja==1.11.1.1
|
||||
nltk==3.8.1
|
||||
notebook==7.0.3
|
||||
notebook==7.0.6
|
||||
notebook_shim==0.2.3
|
||||
numpy==1.25.2
|
||||
nvidia-cublas-cu11==11.10.3.66
|
||||
nvidia-cuda-cupti-cu11==11.7.101
|
||||
nvidia-cuda-nvrtc-cu11==11.7.99
|
||||
nvidia-cuda-runtime-cu11==11.7.99
|
||||
nvidia-cudnn-cu11==8.5.0.96
|
||||
nvidia-cufft-cu11==10.9.0.58
|
||||
nvidia-curand-cu11==10.2.10.91
|
||||
nvidia-cusolver-cu11==11.4.0.1
|
||||
nvidia-cusparse-cu11==11.7.4.91
|
||||
nvidia-nccl-cu11==2.14.3
|
||||
nvidia-nvtx-cu11==11.7.91
|
||||
openai==0.28.0
|
||||
numpy==1.26.1
|
||||
nvidia-cublas-cu12==12.1.3.1
|
||||
nvidia-cuda-cupti-cu12==12.1.105
|
||||
nvidia-cuda-nvrtc-cu12==12.1.105
|
||||
nvidia-cuda-runtime-cu12==12.1.105
|
||||
nvidia-cudnn-cu12==8.9.2.26
|
||||
nvidia-cufft-cu12==11.0.2.54
|
||||
nvidia-curand-cu12==10.3.2.106
|
||||
nvidia-cusolver-cu12==11.4.5.107
|
||||
nvidia-cusparse-cu12==12.1.0.106
|
||||
nvidia-nccl-cu12==2.18.1
|
||||
nvidia-nvjitlink-cu12==12.3.52
|
||||
nvidia-nvtx-cu12==12.1.105
|
||||
openai==1.2.3
|
||||
overrides==7.4.0
|
||||
packaging==23.1
|
||||
pandas==2.1.0
|
||||
packaging==23.2
|
||||
pandas==2.1.3
|
||||
pandocfilters==1.5.0
|
||||
papermill==2.4.0
|
||||
papermill==2.5.0
|
||||
paramiko==3.3.1
|
||||
parso==0.8.3
|
||||
pathspec==0.11.2
|
||||
pathtools==0.1.2
|
||||
patsy==0.5.3
|
||||
peft==0.5.0
|
||||
peft==0.6.1
|
||||
pexpect==4.8.0
|
||||
pickleshare==0.7.5
|
||||
Pillow==10.0.0
|
||||
platformdirs==3.10.0
|
||||
plotly==5.18.0
|
||||
pluggy==1.3.0
|
||||
prometheus-client==0.17.1
|
||||
prompt-toolkit==3.0.39
|
||||
promptsource @ git+https://github.com/bigscience-workshop/promptsource@7dab96a3eeb3717cea633705135ebc488885d709
|
||||
Pillow==10.1.0
|
||||
pipreqs==0.4.13
|
||||
platformdirs==4.0.0
|
||||
prometheus-client==0.18.0
|
||||
prompt-toolkit==3.0.40
|
||||
protobuf==3.20.3
|
||||
psutil==5.9.5
|
||||
psutil==5.9.6
|
||||
ptyprocess==0.7.0
|
||||
pure-eval==0.2.2
|
||||
py-cpuinfo==9.0.0
|
||||
py7zr==0.20.6
|
||||
pyarrow==13.0.0
|
||||
pyasn1==0.5.0
|
||||
pybcj==1.0.1
|
||||
pycodestyle==2.11.1
|
||||
pyarrow==14.0.1
|
||||
pycparser==2.21
|
||||
pycryptodomex==3.19.0
|
||||
pydantic==1.10.12
|
||||
pydeck==0.8.1b0
|
||||
pyflakes==3.1.0
|
||||
pydantic==2.4.2
|
||||
pydantic_core==2.10.1
|
||||
Pygments==2.16.1
|
||||
PyNaCl==1.5.0
|
||||
pyparsing==3.0.9
|
||||
pyppmd==1.0.0
|
||||
pytest==7.4.3
|
||||
pynvml==11.5.0
|
||||
pyparsing==3.1.1
|
||||
python-dateutil==2.8.2
|
||||
python-json-logger==2.0.7
|
||||
pytz==2023.3.post1
|
||||
PyYAML==6.0.1
|
||||
pyzmq==25.1.1
|
||||
pyzstd==0.15.9
|
||||
qtconsole==5.4.4
|
||||
QtPy==2.4.0
|
||||
qtconsole==5.5.0
|
||||
QtPy==2.4.1
|
||||
referencing==0.30.2
|
||||
regex==2023.8.8
|
||||
regex==2023.10.3
|
||||
requests==2.31.0
|
||||
rfc3339-validator==0.1.4
|
||||
rfc3986-validator==0.1.1
|
||||
rich==13.6.0
|
||||
rouge-score==0.1.2
|
||||
rpds-py==0.10.2
|
||||
rsa==4.7.2
|
||||
rpds-py==0.12.0
|
||||
s3transfer==0.7.0
|
||||
safetensors==0.3.3
|
||||
scikit-learn==1.3.0
|
||||
scipy==1.11.2
|
||||
seaborn==0.12.2
|
||||
safetensors==0.4.0
|
||||
scikit-learn==1.3.2
|
||||
scipy==1.11.3
|
||||
seaborn==0.13.0
|
||||
Send2Trash==1.8.2
|
||||
sentencepiece==0.1.99
|
||||
sentry-sdk==1.30.0
|
||||
setproctitle==1.3.2
|
||||
sentry-sdk==1.34.0
|
||||
setproctitle==1.3.3
|
||||
shtab==1.6.4
|
||||
six==1.16.0
|
||||
smmap==5.0.0
|
||||
smmap==5.0.1
|
||||
sniffio==1.3.0
|
||||
soupsieve==2.5
|
||||
stack-data==0.6.2
|
||||
statsmodels==0.14.0
|
||||
streamlit==0.82.0
|
||||
stack-data==0.6.3
|
||||
sympy==1.12
|
||||
tenacity==8.2.3
|
||||
termcolor==2.3.0
|
||||
terminado==0.17.1
|
||||
texttable==1.7.0
|
||||
textwrap3==0.9.2
|
||||
terminado==0.18.0
|
||||
threadpoolctl==3.2.0
|
||||
tinycss2==1.2.1
|
||||
tokenizers==0.13.3
|
||||
toml==0.10.2
|
||||
tomli==1.2.3
|
||||
toolz==0.12.0
|
||||
torch==2.0.1
|
||||
torchvision==0.15.2
|
||||
tokenizers==0.14.1
|
||||
tomli==2.0.1
|
||||
torch==2.1.0
|
||||
tornado==6.3.3
|
||||
tqdm==4.66.1
|
||||
traitlets==5.9.0
|
||||
transformers==4.33.1
|
||||
triton==2.0.0
|
||||
trl==0.7.1
|
||||
typing_extensions==4.7.1
|
||||
traitlets==5.13.0
|
||||
transformers==4.35.0
|
||||
triton==2.1.0
|
||||
trl==0.7.4
|
||||
types-python-dateutil==2.8.19.14
|
||||
typing_extensions==4.8.0
|
||||
tyro==0.5.12
|
||||
tzdata==2023.3
|
||||
tzlocal==5.2
|
||||
uri-template==1.3.0
|
||||
urllib3==1.26.16
|
||||
validators==0.22.0
|
||||
wandb==0.15.11
|
||||
watchdog==3.0.0
|
||||
wcwidth==0.2.6
|
||||
urllib3==2.0.7
|
||||
wandb==0.16.0
|
||||
wcwidth==0.2.9
|
||||
webcolors==1.13
|
||||
webencodings==0.5.1
|
||||
websocket-client==1.6.2
|
||||
widgetsnbextension==4.0.8
|
||||
xxhash==3.3.0
|
||||
websocket-client==1.6.4
|
||||
widgetsnbextension==4.0.9
|
||||
xxhash==3.4.1
|
||||
yarg==0.1.9
|
||||
yarl==1.9.2
|
||||
|
||||
@@ -1,7 +1,9 @@
|
||||
import setuptools
|
||||
import os
|
||||
from io import open
|
||||
from setuptools.command.install import install
|
||||
from setuptools import find_packages
|
||||
from download_data import download_data
|
||||
|
||||
src_dir = os.path.abspath(os.path.dirname(__file__))
|
||||
|
||||
@@ -15,24 +17,35 @@ if os.path.isfile(requirements_path):
|
||||
with open(requirements_path) as f:
|
||||
requirements = f.read().splitlines()
|
||||
|
||||
class PostInstallCommand(install):
|
||||
"""Post-installation for installation mode."""
|
||||
def run(self):
|
||||
print("RUNNING POST INSTALL 1")
|
||||
install.run(self)
|
||||
print("RUNNING POST INSTALL")
|
||||
download_data()
|
||||
|
||||
|
||||
setuptools.setup(
|
||||
name="fig-benchmark",
|
||||
name="genies-benchmark",
|
||||
version="0.0.1",
|
||||
author="Joshua Clymer, Garrett Baker, Rohan Subramani, and Sam Wang",
|
||||
author_email="joshuamclymer@gmail.com",
|
||||
description="The fig benchmark repository contains datasets and tooling for evaluating the generalization of preferrence models.",
|
||||
long_description=long_description,
|
||||
long_description_content_type="text/markdown",
|
||||
url="https://github.com/Joshuaclymer/FIG-benchmark",
|
||||
url="https://github.com/Joshuaclymer/GENIES",
|
||||
classifiers=[
|
||||
"Programming Language :: Python :: 3",
|
||||
"License :: MIT License",
|
||||
"Operating System :: OS Independent",
|
||||
],
|
||||
cmdclass={
|
||||
'install': PostInstallCommand,
|
||||
},
|
||||
python_requires='>=3.10',
|
||||
install_requires=requirements,
|
||||
packages=find_packages(where='src'), # Specify 'src' as the root
|
||||
package_dir={'': 'src'},
|
||||
package_data={'fig-benchmark': ['LICENCE', 'requirements.txt']},
|
||||
package_data={'genies-benchmark': ['LICENCE', 'requirements.txt']},
|
||||
)
|
||||
@@ -0,0 +1,52 @@
|
||||
Metadata-Version: 2.1
|
||||
Name: fig-benchmark
|
||||
Version: 0.0.1
|
||||
Summary: The fig benchmark repository contains datasets and tooling for evaluating the generalization of preferrence models.
|
||||
Home-page: https://github.com/Joshuaclymer/FIG-benchmark
|
||||
Author: Joshua Clymer, Garrett Baker, Rohan Subramani, and Sam Wang
|
||||
Author-email: joshuamclymer@gmail.com
|
||||
Classifier: Programming Language :: Python :: 3
|
||||
Classifier: License :: MIT License
|
||||
Classifier: Operating System :: OS Independent
|
||||
Requires-Python: >=3.10
|
||||
Description-Content-Type: text/markdown
|
||||
|
||||
# Generalization Analogies: A Testbed for Generalizing AI Oversight to Hard-To-Measure Domains
|
||||
Read our paper [here](TODO). Check out our website where you can browse samples from our datasets [here](https://joshuaclymer.github.io/generalization-analogies-website/).
|
||||
|
||||

|
||||
## Abstract
|
||||
As AI systems become more capable and are deployed in complex environments, it may become challenging to verify that they follow instructions; however, the limitations of human oversight could be overcome by controlling how LLMs generalize human feedback to contexts where it is unreliable. To better understand how Reward Models generalize human feedback, we craft 69 distribution shifts spanning 8 different categories. We find that Reward Models do not learn to evaluate instruction-following by default and instead favor personas that resemble internet text. Techniques for interpreting Reward Model’s internal representations achieve better generalization, but still frequently fail to distinguish instruction-following from conflated behaviors. We consolidate the 15 most challenging distribution shifts into the \textbf{GEN}aralization analog\textbf{IES} (\textsc{GENIES}) benchmark, which we hope will enable progress toward controlling Reward Model generalization.
|
||||
|
||||
## Quickstart
|
||||
|
||||
This repository contains:
|
||||
- Our datasets (`./distributions`) along with pairing specifications (`./distribution_shifts`).
|
||||
- Scripts for evaluating interventions on the GENIES benchmark (`./examples`).
|
||||
- Our results (`./results`).
|
||||
- Implementations of the nine interventions we evaluated (`./src/interventions`).
|
||||
|
||||
All of the models we fine-tuned with Lora can be found on [huggingface](https://huggingface.co/genies-models).
|
||||
|
||||
**Setup:**
|
||||
```
|
||||
conda create --name env python=3.10
|
||||
pip install -e .
|
||||
python download_model_from_hf.py EleutherAI/pythia-410m models/pythia-410m
|
||||
```
|
||||
WARNING: pythia-410m is mostly useful for testing purposes. Most tuning interventions perform poorly with this model.
|
||||
|
||||
## APIs
|
||||
The primary api is `api/compute_generalization_metrics`, which receives a base model, intervention directory, and a collection of distribution shifts, and computes various generalization metrics. See `examples/compute_generalization_metrics.sh` for example usage.
|
||||
|
||||
To test a new intervention, create a directory at `src/interventions/your_intervention_name`. This directory must contain a `train.py` file and an `eval.py` file.
|
||||
|
||||
`src/interventions/your_intervention_name/train.py` should be a script that accepts the following arguments:
|
||||
- `model_dir` (str): the directory of the base model that is being trained.
|
||||
- `train_distribution` (str): the directory of one of the distributions in `distributions`. For example: `distributions/alpaca_mmlu`.
|
||||
- `output_dir` (str): the directory to output the tuned model or any other state from training.
|
||||
|
||||
`src/interventions/your_intervention_name/eval.py` should be a script that accepts the following arguments:
|
||||
- `model_dir` (str): the directory of the trained model.
|
||||
- `distribution_dirs` (List\[str\]): a list of subdirectories of `distributions`.
|
||||
- `output_paths` (List\[str\]): where to save the results. The results should be json files. The only required key is `eval_accuracy`. Evaluation results are stored in `results/evaluations`.# GENIES
|
||||
@@ -0,0 +1,15 @@
|
||||
README.md
|
||||
setup.py
|
||||
src/api/__init__.py
|
||||
src/api/compute_generalization_metrics.py
|
||||
src/api/data_classes.py
|
||||
src/api/evaluate.py
|
||||
src/api/hyperparameter_sweep.py
|
||||
src/api/model.py
|
||||
src/api/train.py
|
||||
src/api/util.py
|
||||
src/fig_benchmark.egg-info/PKG-INFO
|
||||
src/fig_benchmark.egg-info/SOURCES.txt
|
||||
src/fig_benchmark.egg-info/dependency_links.txt
|
||||
src/fig_benchmark.egg-info/requires.txt
|
||||
src/fig_benchmark.egg-info/top_level.txt
|
||||
@@ -0,0 +1 @@
|
||||
|
||||
@@ -0,0 +1,203 @@
|
||||
absl-py==2.0.0
|
||||
accelerate==0.24.1
|
||||
aiohttp==3.8.6
|
||||
aiosignal==1.3.1
|
||||
annotated-types==0.6.0
|
||||
anyio==3.7.1
|
||||
appdirs==1.4.4
|
||||
argon2-cffi==23.1.0
|
||||
argon2-cffi-bindings==21.2.0
|
||||
arrow==1.3.0
|
||||
asttokens==2.4.1
|
||||
async-lru==2.0.4
|
||||
async-timeout==4.0.3
|
||||
attrs==23.1.0
|
||||
Babel==2.13.1
|
||||
bcrypt==4.0.1
|
||||
beautifulsoup4==4.12.2
|
||||
bitsandbytes==0.41.2.post2
|
||||
bleach==6.1.0
|
||||
boto3==1.28.84
|
||||
botocore==1.31.84
|
||||
certifi==2023.7.22
|
||||
cffi==1.16.0
|
||||
charset-normalizer==3.3.2
|
||||
click==8.1.7
|
||||
comm==0.2.0
|
||||
contourpy==1.2.0
|
||||
cryptography==41.0.5
|
||||
cycler==0.12.1
|
||||
Cython==0.29.36
|
||||
datasets==2.14.6
|
||||
debugpy==1.8.0
|
||||
decorator==5.1.1
|
||||
deepspeed==0.12.2
|
||||
defusedxml==0.7.1
|
||||
dill==0.3.7
|
||||
distro==1.8.0
|
||||
docker-pycreds==0.4.0
|
||||
docopt==0.6.2
|
||||
docstring-parser==0.15
|
||||
einops==0.7.0
|
||||
entrypoints==0.4
|
||||
exceptiongroup==1.1.3
|
||||
executing==2.0.1
|
||||
fastjsonschema==2.18.1
|
||||
filelock==3.13.1
|
||||
fire==0.5.0
|
||||
fonttools==4.44.0
|
||||
fqdn==1.5.1
|
||||
frozenlist==1.4.0
|
||||
fsspec==2023.10.0
|
||||
gitdb==4.0.11
|
||||
GitPython==3.1.40
|
||||
h11==0.14.0
|
||||
hjson==3.1.0
|
||||
httpcore==1.0.2
|
||||
httpx==0.25.1
|
||||
huggingface-hub==0.17.3
|
||||
idna==3.4
|
||||
ipykernel==6.26.0
|
||||
ipython==8.17.2
|
||||
ipywidgets==8.1.1
|
||||
isoduration==20.11.0
|
||||
jedi==0.19.1
|
||||
Jinja2==3.1.2
|
||||
jmespath==1.0.1
|
||||
joblib==1.3.2
|
||||
json5==0.9.14
|
||||
jsonpointer==2.4
|
||||
jsonschema==4.19.2
|
||||
jsonschema-specifications==2023.7.1
|
||||
jupyter==1.0.0
|
||||
jupyter-console==6.6.3
|
||||
jupyter-events==0.9.0
|
||||
jupyter-lsp==2.2.0
|
||||
jupyter_client==8.6.0
|
||||
jupyter_core==5.5.0
|
||||
jupyter_server==2.10.0
|
||||
jupyter_server_terminals==0.4.4
|
||||
jupyterlab==4.0.8
|
||||
jupyterlab-pygments==0.2.2
|
||||
jupyterlab-widgets==3.0.9
|
||||
jupyterlab_server==2.25.1
|
||||
kiwisolver==1.4.5
|
||||
markdown-it-py==3.0.0
|
||||
MarkupSafe==2.1.3
|
||||
matplotlib==3.8.1
|
||||
matplotlib-inline==0.1.6
|
||||
mdurl==0.1.2
|
||||
mistune==3.0.2
|
||||
mpmath==1.3.0
|
||||
multidict==6.0.4
|
||||
multiprocess==0.70.15
|
||||
nbclient==0.9.0
|
||||
nbconvert==7.11.0
|
||||
nbformat==5.9.2
|
||||
nest-asyncio==1.5.8
|
||||
networkx==3.2.1
|
||||
ninja==1.11.1.1
|
||||
nltk==3.8.1
|
||||
notebook==7.0.6
|
||||
notebook_shim==0.2.3
|
||||
numpy==1.26.1
|
||||
nvidia-cublas-cu12==12.1.3.1
|
||||
nvidia-cuda-cupti-cu12==12.1.105
|
||||
nvidia-cuda-nvrtc-cu12==12.1.105
|
||||
nvidia-cuda-runtime-cu12==12.1.105
|
||||
nvidia-cudnn-cu12==8.9.2.26
|
||||
nvidia-cufft-cu12==11.0.2.54
|
||||
nvidia-curand-cu12==10.3.2.106
|
||||
nvidia-cusolver-cu12==11.4.5.107
|
||||
nvidia-cusparse-cu12==12.1.0.106
|
||||
nvidia-nccl-cu12==2.18.1
|
||||
nvidia-nvjitlink-cu12==12.3.52
|
||||
nvidia-nvtx-cu12==12.1.105
|
||||
openai==1.2.3
|
||||
overrides==7.4.0
|
||||
packaging==23.2
|
||||
pandas==2.1.3
|
||||
pandocfilters==1.5.0
|
||||
papermill==2.5.0
|
||||
paramiko==3.3.1
|
||||
parso==0.8.3
|
||||
peft==0.6.1
|
||||
pexpect==4.8.0
|
||||
Pillow==10.1.0
|
||||
pipreqs==0.4.13
|
||||
platformdirs==4.0.0
|
||||
prometheus-client==0.18.0
|
||||
prompt-toolkit==3.0.40
|
||||
protobuf==3.20.3
|
||||
psutil==5.9.6
|
||||
ptyprocess==0.7.0
|
||||
pure-eval==0.2.2
|
||||
py-cpuinfo==9.0.0
|
||||
pyarrow==14.0.1
|
||||
pycparser==2.21
|
||||
pydantic==2.4.2
|
||||
pydantic_core==2.10.1
|
||||
Pygments==2.16.1
|
||||
PyNaCl==1.5.0
|
||||
pynvml==11.5.0
|
||||
pyparsing==3.1.1
|
||||
python-dateutil==2.8.2
|
||||
python-json-logger==2.0.7
|
||||
pytz==2023.3.post1
|
||||
PyYAML==6.0.1
|
||||
pyzmq==25.1.1
|
||||
qtconsole==5.5.0
|
||||
QtPy==2.4.1
|
||||
referencing==0.30.2
|
||||
regex==2023.10.3
|
||||
requests==2.31.0
|
||||
rfc3339-validator==0.1.4
|
||||
rfc3986-validator==0.1.1
|
||||
rich==13.6.0
|
||||
rouge-score==0.1.2
|
||||
rpds-py==0.12.0
|
||||
s3transfer==0.7.0
|
||||
safetensors==0.4.0
|
||||
scikit-learn==1.3.2
|
||||
scipy==1.11.3
|
||||
seaborn==0.13.0
|
||||
Send2Trash==1.8.2
|
||||
sentencepiece==0.1.99
|
||||
sentry-sdk==1.34.0
|
||||
setproctitle==1.3.3
|
||||
shtab==1.6.4
|
||||
six==1.16.0
|
||||
smmap==5.0.1
|
||||
sniffio==1.3.0
|
||||
soupsieve==2.5
|
||||
stack-data==0.6.3
|
||||
sympy==1.12
|
||||
tenacity==8.2.3
|
||||
termcolor==2.3.0
|
||||
terminado==0.18.0
|
||||
threadpoolctl==3.2.0
|
||||
tinycss2==1.2.1
|
||||
tokenizers==0.14.1
|
||||
tomli==2.0.1
|
||||
torch==2.1.0
|
||||
tornado==6.3.3
|
||||
tqdm==4.66.1
|
||||
traitlets==5.13.0
|
||||
transformers==4.35.0
|
||||
triton==2.1.0
|
||||
trl==0.7.4
|
||||
types-python-dateutil==2.8.19.14
|
||||
typing_extensions==4.8.0
|
||||
tyro==0.5.12
|
||||
tzdata==2023.3
|
||||
uri-template==1.3.0
|
||||
urllib3==2.0.7
|
||||
wandb==0.16.0
|
||||
wcwidth==0.2.9
|
||||
webcolors==1.13
|
||||
webencodings==0.5.1
|
||||
websocket-client==1.6.4
|
||||
widgetsnbextension==4.0.9
|
||||
xxhash==3.4.1
|
||||
yarg==0.1.9
|
||||
yarl==1.9.2
|
||||
@@ -0,0 +1 @@
|
||||
api
|
||||
@@ -0,0 +1,52 @@
|
||||
Metadata-Version: 2.1
|
||||
Name: genies-benchmark
|
||||
Version: 0.0.1
|
||||
Summary: The fig benchmark repository contains datasets and tooling for evaluating the generalization of preferrence models.
|
||||
Home-page: https://github.com/Joshuaclymer/GENIES
|
||||
Author: Joshua Clymer, Garrett Baker, Rohan Subramani, and Sam Wang
|
||||
Author-email: joshuamclymer@gmail.com
|
||||
Classifier: Programming Language :: Python :: 3
|
||||
Classifier: License :: MIT License
|
||||
Classifier: Operating System :: OS Independent
|
||||
Requires-Python: >=3.10
|
||||
Description-Content-Type: text/markdown
|
||||
|
||||
# Generalization Analogies: A Testbed for Generalizing AI Oversight to Hard-To-Measure Domains
|
||||
Read our paper [here](TODO). Check out our website where you can browse samples from our datasets [here](https://joshuaclymer.github.io/generalization-analogies-website/).
|
||||
|
||||

|
||||
## Abstract
|
||||
As AI systems become more capable and are deployed in complex environments, it may become challenging to verify that they follow instructions; however, the limitations of human oversight could be overcome by controlling how LLMs generalize human feedback to contexts where it is unreliable. To better understand how Reward Models generalize human feedback, we craft 69 distribution shifts spanning 8 different categories. We find that Reward Models do not learn to evaluate instruction-following by default and instead favor personas that resemble internet text. Techniques for interpreting Reward Model’s internal representations achieve better generalization, but still frequently fail to distinguish instruction-following from conflated behaviors. We consolidate the 15 most challenging distribution shifts into the \textbf{GEN}aralization analog\textbf{IES} (\textsc{GENIES}) benchmark, which we hope will enable progress toward controlling Reward Model generalization.
|
||||
|
||||
## Quickstart
|
||||
|
||||
This repository contains:
|
||||
- Our datasets (`./distributions`) along with pairing specifications (`./distribution_shifts`). Download our datasets [here](https://genies-data.s3.us-east-2.amazonaws.com/genies-datasets.tar) or run the setup command after cloning the repo.
|
||||
- Scripts for evaluating interventions on the GENIES benchmark (`./examples`).
|
||||
- Our results (`./results`).
|
||||
- Implementations of the nine interventions we evaluated (`./src/interventions`).
|
||||
|
||||
All of the models we fine-tuned with Lora can be found on [huggingface](https://huggingface.co/genies-models).
|
||||
|
||||
**Setup:**
|
||||
```
|
||||
conda create --name env python=3.10
|
||||
pip install .
|
||||
python download_model_from_hf.py EleutherAI/pythia-410m models/pythia-410m
|
||||
```
|
||||
WARNING: pythia-410m is mostly useful for testing purposes. Most tuning interventions perform poorly with this model.
|
||||
|
||||
## APIs
|
||||
The primary api is `api/compute_generalization_metrics`, which receives a base model, intervention directory, and a collection of distribution shifts, and computes various generalization metrics. See `examples/compute_generalization_metrics.sh` for example usage.
|
||||
|
||||
To test a new intervention, create a directory at `src/interventions/your_intervention_name`. This directory must contain a `train.py` file and an `eval.py` file.
|
||||
|
||||
`src/interventions/your_intervention_name/train.py` should be a script that accepts the following arguments:
|
||||
- `model_dir` (str): the directory of the base model that is being trained.
|
||||
- `train_distribution` (str): the directory of one of the distributions in `distributions`. For example: `distributions/alpaca_mmlu`.
|
||||
- `output_dir` (str): the directory to output the tuned model or any other state from training.
|
||||
|
||||
`src/interventions/your_intervention_name/eval.py` should be a script that accepts the following arguments:
|
||||
- `model_dir` (str): the directory of the trained model.
|
||||
- `distribution_dirs` (List\[str\]): a list of subdirectories of `distributions`.
|
||||
- `output_paths` (List\[str\]): where to save the results. The results should be json files. The only required key is `eval_accuracy`. Evaluation results are stored in `results/evaluations`.# GENIES
|
||||
@@ -0,0 +1,15 @@
|
||||
README.md
|
||||
setup.py
|
||||
src/api/__init__.py
|
||||
src/api/compute_generalization_metrics.py
|
||||
src/api/data_classes.py
|
||||
src/api/evaluate.py
|
||||
src/api/hyperparameter_sweep.py
|
||||
src/api/model.py
|
||||
src/api/train.py
|
||||
src/api/util.py
|
||||
src/genies_benchmark.egg-info/PKG-INFO
|
||||
src/genies_benchmark.egg-info/SOURCES.txt
|
||||
src/genies_benchmark.egg-info/dependency_links.txt
|
||||
src/genies_benchmark.egg-info/requires.txt
|
||||
src/genies_benchmark.egg-info/top_level.txt
|
||||
@@ -0,0 +1 @@
|
||||
|
||||
@@ -0,0 +1,203 @@
|
||||
absl-py==2.0.0
|
||||
accelerate==0.24.1
|
||||
aiohttp==3.8.6
|
||||
aiosignal==1.3.1
|
||||
annotated-types==0.6.0
|
||||
anyio==3.7.1
|
||||
appdirs==1.4.4
|
||||
argon2-cffi==23.1.0
|
||||
argon2-cffi-bindings==21.2.0
|
||||
arrow==1.3.0
|
||||
asttokens==2.4.1
|
||||
async-lru==2.0.4
|
||||
async-timeout==4.0.3
|
||||
attrs==23.1.0
|
||||
Babel==2.13.1
|
||||
bcrypt==4.0.1
|
||||
beautifulsoup4==4.12.2
|
||||
bitsandbytes==0.41.2.post2
|
||||
bleach==6.1.0
|
||||
boto3==1.28.84
|
||||
botocore==1.31.84
|
||||
certifi==2023.7.22
|
||||
cffi==1.16.0
|
||||
charset-normalizer==3.3.2
|
||||
click==8.1.7
|
||||
comm==0.2.0
|
||||
contourpy==1.2.0
|
||||
cryptography==41.0.5
|
||||
cycler==0.12.1
|
||||
Cython==0.29.36
|
||||
datasets==2.14.6
|
||||
debugpy==1.8.0
|
||||
decorator==5.1.1
|
||||
deepspeed==0.12.2
|
||||
defusedxml==0.7.1
|
||||
dill==0.3.7
|
||||
distro==1.8.0
|
||||
docker-pycreds==0.4.0
|
||||
docopt==0.6.2
|
||||
docstring-parser==0.15
|
||||
einops==0.7.0
|
||||
entrypoints==0.4
|
||||
exceptiongroup==1.1.3
|
||||
executing==2.0.1
|
||||
fastjsonschema==2.18.1
|
||||
filelock==3.13.1
|
||||
fire==0.5.0
|
||||
fonttools==4.44.0
|
||||
fqdn==1.5.1
|
||||
frozenlist==1.4.0
|
||||
fsspec==2023.10.0
|
||||
gitdb==4.0.11
|
||||
GitPython==3.1.40
|
||||
h11==0.14.0
|
||||
hjson==3.1.0
|
||||
httpcore==1.0.2
|
||||
httpx==0.25.1
|
||||
huggingface-hub==0.17.3
|
||||
idna==3.4
|
||||
ipykernel==6.26.0
|
||||
ipython==8.17.2
|
||||
ipywidgets==8.1.1
|
||||
isoduration==20.11.0
|
||||
jedi==0.19.1
|
||||
Jinja2==3.1.2
|
||||
jmespath==1.0.1
|
||||
joblib==1.3.2
|
||||
json5==0.9.14
|
||||
jsonpointer==2.4
|
||||
jsonschema==4.19.2
|
||||
jsonschema-specifications==2023.7.1
|
||||
jupyter==1.0.0
|
||||
jupyter-console==6.6.3
|
||||
jupyter-events==0.9.0
|
||||
jupyter-lsp==2.2.0
|
||||
jupyter_client==8.6.0
|
||||
jupyter_core==5.5.0
|
||||
jupyter_server==2.10.0
|
||||
jupyter_server_terminals==0.4.4
|
||||
jupyterlab==4.0.8
|
||||
jupyterlab-pygments==0.2.2
|
||||
jupyterlab-widgets==3.0.9
|
||||
jupyterlab_server==2.25.1
|
||||
kiwisolver==1.4.5
|
||||
markdown-it-py==3.0.0
|
||||
MarkupSafe==2.1.3
|
||||
matplotlib==3.8.1
|
||||
matplotlib-inline==0.1.6
|
||||
mdurl==0.1.2
|
||||
mistune==3.0.2
|
||||
mpmath==1.3.0
|
||||
multidict==6.0.4
|
||||
multiprocess==0.70.15
|
||||
nbclient==0.9.0
|
||||
nbconvert==7.11.0
|
||||
nbformat==5.9.2
|
||||
nest-asyncio==1.5.8
|
||||
networkx==3.2.1
|
||||
ninja==1.11.1.1
|
||||
nltk==3.8.1
|
||||
notebook==7.0.6
|
||||
notebook_shim==0.2.3
|
||||
numpy==1.26.1
|
||||
nvidia-cublas-cu12==12.1.3.1
|
||||
nvidia-cuda-cupti-cu12==12.1.105
|
||||
nvidia-cuda-nvrtc-cu12==12.1.105
|
||||
nvidia-cuda-runtime-cu12==12.1.105
|
||||
nvidia-cudnn-cu12==8.9.2.26
|
||||
nvidia-cufft-cu12==11.0.2.54
|
||||
nvidia-curand-cu12==10.3.2.106
|
||||
nvidia-cusolver-cu12==11.4.5.107
|
||||
nvidia-cusparse-cu12==12.1.0.106
|
||||
nvidia-nccl-cu12==2.18.1
|
||||
nvidia-nvjitlink-cu12==12.3.52
|
||||
nvidia-nvtx-cu12==12.1.105
|
||||
openai==1.2.3
|
||||
overrides==7.4.0
|
||||
packaging==23.2
|
||||
pandas==2.1.3
|
||||
pandocfilters==1.5.0
|
||||
papermill==2.5.0
|
||||
paramiko==3.3.1
|
||||
parso==0.8.3
|
||||
peft==0.6.1
|
||||
pexpect==4.8.0
|
||||
Pillow==10.1.0
|
||||
pipreqs==0.4.13
|
||||
platformdirs==4.0.0
|
||||
prometheus-client==0.18.0
|
||||
prompt-toolkit==3.0.40
|
||||
protobuf==3.20.3
|
||||
psutil==5.9.6
|
||||
ptyprocess==0.7.0
|
||||
pure-eval==0.2.2
|
||||
py-cpuinfo==9.0.0
|
||||
pyarrow==14.0.1
|
||||
pycparser==2.21
|
||||
pydantic==2.4.2
|
||||
pydantic_core==2.10.1
|
||||
Pygments==2.16.1
|
||||
PyNaCl==1.5.0
|
||||
pynvml==11.5.0
|
||||
pyparsing==3.1.1
|
||||
python-dateutil==2.8.2
|
||||
python-json-logger==2.0.7
|
||||
pytz==2023.3.post1
|
||||
PyYAML==6.0.1
|
||||
pyzmq==25.1.1
|
||||
qtconsole==5.5.0
|
||||
QtPy==2.4.1
|
||||
referencing==0.30.2
|
||||
regex==2023.10.3
|
||||
requests==2.31.0
|
||||
rfc3339-validator==0.1.4
|
||||
rfc3986-validator==0.1.1
|
||||
rich==13.6.0
|
||||
rouge-score==0.1.2
|
||||
rpds-py==0.12.0
|
||||
s3transfer==0.7.0
|
||||
safetensors==0.4.0
|
||||
scikit-learn==1.3.2
|
||||
scipy==1.11.3
|
||||
seaborn==0.13.0
|
||||
Send2Trash==1.8.2
|
||||
sentencepiece==0.1.99
|
||||
sentry-sdk==1.34.0
|
||||
setproctitle==1.3.3
|
||||
shtab==1.6.4
|
||||
six==1.16.0
|
||||
smmap==5.0.1
|
||||
sniffio==1.3.0
|
||||
soupsieve==2.5
|
||||
stack-data==0.6.3
|
||||
sympy==1.12
|
||||
tenacity==8.2.3
|
||||
termcolor==2.3.0
|
||||
terminado==0.18.0
|
||||
threadpoolctl==3.2.0
|
||||
tinycss2==1.2.1
|
||||
tokenizers==0.14.1
|
||||
tomli==2.0.1
|
||||
torch==2.1.0
|
||||
tornado==6.3.3
|
||||
tqdm==4.66.1
|
||||
traitlets==5.13.0
|
||||
transformers==4.35.0
|
||||
triton==2.1.0
|
||||
trl==0.7.4
|
||||
types-python-dateutil==2.8.19.14
|
||||
typing_extensions==4.8.0
|
||||
tyro==0.5.12
|
||||
tzdata==2023.3
|
||||
uri-template==1.3.0
|
||||
urllib3==2.0.7
|
||||
wandb==0.16.0
|
||||
wcwidth==0.2.9
|
||||
webcolors==1.13
|
||||
webencodings==0.5.1
|
||||
websocket-client==1.6.4
|
||||
widgetsnbextension==4.0.9
|
||||
xxhash==3.4.1
|
||||
yarg==0.1.9
|
||||
yarl==1.9.2
|
||||
@@ -0,0 +1 @@
|
||||
api
|
||||
@@ -0,0 +1,29 @@
|
||||
import fire
|
||||
import os
|
||||
import api.util as util
|
||||
import json
|
||||
import boto3
|
||||
import os
|
||||
import tarfile
|
||||
|
||||
def make_tar(directory_path, output_filename):
|
||||
"""Compresses directory into a tar file."""
|
||||
with tarfile.open(output_filename, "w") as tar:
|
||||
tar.add(directory_path, arcname=os.path.basename(directory_path))
|
||||
|
||||
def upload_directory_to_s3(directory_path, bucket_name, tar_name):
|
||||
make_tar(directory_path, tar_name)
|
||||
|
||||
# Upload to S3
|
||||
s3 = boto3.client('s3')
|
||||
s3.upload_file(tar_name, bucket_name, tar_name)
|
||||
|
||||
print(f"Uploaded {tar_name} to {bucket_name}")
|
||||
print("The link is: ", f"https://genies-data.s3.us-east-2.amazonaws.com/genies-datasets.tar")
|
||||
|
||||
|
||||
def upload_data():
|
||||
upload_directory_to_s3(f"distributions", "genies-data", f"genies-datasets.tar")
|
||||
|
||||
if __name__ == "__main__":
|
||||
fire.Fire(upload_data)
|
||||
Reference in New Issue
Block a user