Install using the setuptools.build_meta backend

pyproject.toml

@@ -1,3 +1,7 @@
+[build-system]
+requires = ["setuptools"]
+build-backend = "setuptools.build_meta"
+
 # Settings for Black: The Uncompromising Code Formatter
 [tool.black]
-line-length = 120
+line-length = 120

setup.cfg

@@ -1,3 +1,103 @@
+[metadata]
+name = crfm-helm
+version = 0.2.2
+author = Stanford CRFM
+author_email = contact-crfm@stanford.edu
+description = Benchmark for language models
+long_description = Benchmark for language models
+keywords = language models benchmarking
+license = Apache License 2.0
+classifiers =
+    Programming Language :: Python :: 3 :: Only
+    Programming Language :: Python :: 3.8
+    License :: OSI Approved :: Apache Software License
+url = "https://github.com/stanford-crfm/helm"
+
+[options]
+python_requires = ~=3.8
+package_dir =
+    =src
+packages = find:
+zip_safe = False
+include_package_data = True
+
+install_requires=
+    # Common
+    zstandard~=0.18.0
+    tqdm~=4.64.1
+    pyhocon~=0.3.59
+    dacite~=1.6.0
+
+    # Proxy
+    aleph-alpha-client~=2.14.0
+    anthropic~=0.2.5
+    bottle~=0.12.23
+    gunicorn~=20.1.0
+    Mako~=1.2.3
+    # sqlitedict==2.0.0 is slow! https://github.com/RaRe-Technologies/sqlitedict/issues/152
+    # Keep the version at 1.7.0.
+    sqlitedict~=1.7.0
+    pymongo~=4.2.0
+    retrying~=1.3.3
+    websocket-client~=1.3.2 # For Anthropic (Legacy stanford-online-all-v4-s3)
+    openai~=0.27.0
+    tiktoken~=0.3.3
+    transformers~=4.28.1
+    tokenizers~=0.13.2
+    icetk~=0.0.4
+    protobuf~=3.20.2  # Can't use 4.21.0 due to backward incompatibility
+    google-api-python-client~=2.64.0
+    revChatGPT~=0.1.1
+
+    # Scenarios
+    datasets~=2.5.2  # For math scenarios
+    gdown~=4.4.0
+    jsonlines~=3.1.0  # Not really needed
+    sympy~=1.11.1  # For math scenarios
+    sentencepiece~=0.1.97
+    numba~=0.56.4
+    cattrs~=22.2.0
+    xlrd~=2.0.1  # Used by pandas.read_excel in ice_scenario
+
+    # Metrics
+    importlib-resources~=5.10.0
+    nltk~=3.7
+    scipy~=1.9.1
+    uncertainty-calibration~=0.1.3
+    rouge-score~=0.1.2
+    pyext~=0.7
+    pytrec_eval==0.5
+    sacrebleu~=2.2.1
+    # Work around https://github.com/p-lambda/verified_calibration/issues/11
+    # TODO: Remove after this issue is resolved
+    scikit-learn~=1.1.2
+    spacy~=3.2.4
+    summ-eval~=0.892
+    surge-api~=1.1.0
+    # End users should install a CUDA version of PyTorch manually if needed
+    torch~=1.12.1  # Summarization metrics
+    torchvision~=0.13.1
+
+    # plotting
+    colorcet~=3.0.1
+    matplotlib~=3.6.0
+    numpy~=1.23.3
+    seaborn~=0.11.0
+
+[options.entry_points]
+console_scripts = 
+    helm-run = helm.benchmark.run:main
+    helm-summarize = helm.benchmark.presentation.summarize:main
+    helm-server = helm.benchmark.server:main
+    helm-create-plots = helm.benchmark.presentation.create_plots:main
+    crfm-proxy-server = helm.proxy.server:main
+    crfm-proxy-cli = helm.proxy.cli:main
+
+[options.packages.find]
+where = src
+exclude =
+    tests*
+
 # Settings for Flake8: Tool For Style Guide Enforcement
 [flake8]
 max-line-length = 120