-
Notifications
You must be signed in to change notification settings - Fork 56
/
setup.cfg
145 lines (123 loc) · 4.08 KB
/
setup.cfg
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
[metadata]
name = decoding-trust
version = 0.1.0
author = Secure Learning Lab
author_email = [email protected]
description = Trustworthiness Benchmark for Large Language Models
long_description = Trustworthiness Benchmark for Large Language Models
keywords = trustworthy language models
license = Creative Commons Public Licenses
classifiers =
Programming Language :: Python :: 3 :: Only
Programming Language :: Python :: 3.9
License :: OSI Approved :: Apache Software License
url = https://github.com/AI-secure/DecodingTrust
[options]
python_requires = >=3.9
package_dir =
=src
packages = find:
zip_safe = False
include_package_data = True
install_requires =
crfm-helm @ git+https://github.com/danielz02/helm.git@main
hydra-core~=1.3.2
fairlearn~=0.9.0
# Common
cattrs~=22.2.0
dacite~=1.6.0
importlib-resources~=5.10.0
Mako~=1.2.3
numpy~=1.23.3
pyhocon~=0.3.59
retrying~=1.3.4
simple-slurm~=0.2.6 # For slurm_jobs
spacy~=3.5.3
tqdm~=4.64.1
zstandard~=0.18.0
# sqlitedict==2.0.0 is slow! https://github.com/RaRe-Technologies/sqlitedict/issues/152
# Keep sqlitedict version at 1.7.0.
sqlitedict~=1.7.0
bottle~=0.12.23
# TODO: Remove these from common
protobuf~=3.20.2 # Can't use 4.21.0 due to backward incompatibility
pymongo~=4.2.0
# Basic Scenarios
datasets~=2.5.2
pyarrow~=11.0.0 # Pinned transitive dependency for datasets; workaround for #1026
jsonlines~=3.1.0 # Not really needed
# Basic metrics
nltk~=3.7
pyext~=0.7
rouge-score~=0.1.2
scipy>=1.9.1,<=1.10.1
uncertainty-calibration~=0.1.3
# Work around https://github.com/p-lambda/verified_calibration/issues/11
# TODO: Remove after this issue is resolved
scikit-learn~=1.1.2
# Model Extras
aleph-alpha-client~=2.14.0
anthropic~=0.2.5
icetk~=0.0.4 # for ice_tokenizer_client
openai~=0.27.8
sentencepiece~=0.1.97 # For palmyra_client and yalm_tokenizer
tiktoken>=0.3.3 # for openai_client
tokenizers~=0.13.3 # for aleph_alpha_client
websocket-client~=1.3.2 # For Anthropic (Legacy stanford-online-all-v4-s3)
# Models and Metrics Extras
transformers>=4.31.0 # For anthropic_client, huggingface_client, huggingface_tokenizer, test_openai_token_cost_estimator, model_summac (via summarization_metrics)
# TODO: Upgrade torch
torch>=1.12.0,<=2.0.1 # For huggingface_client, yalm_tokenizer, model_summac (via summarization_metrics)
torchvision>=0.13.1,<3.0.0 # For huggingface_client, yalm_tokenizer, model_summac (via summarization_metrics)
# Metrics Extras
google-api-python-client~=2.64.0 # For perspective_api_client via toxicity_metrics
# Retry
tenacity~=8.2.2
shortuuid~=1.0.11
colorcet~=3.0.1
matplotlib~=3.6.0
seaborn~=0.11.0
[options.extras_require]
awq =
crfm-helm[awq] @ git+https://github.com/danielz02/helm.git@main
gptq =
torch>=1.12.0,<=2.0.1
crfm-helm[gptq] @ git+https://github.com/danielz02/helm.git@main
quant =
torch>=1.12.0,<=2.0.1
crfm-helm[quant] @ git+https://github.com/danielz02/helm.git@main
slurm =
hydra-submitit-launcher~=1.2.0
submitit~=1.5.0
all =
decoding-trust[quant]
decoding-trust[slurm]
[options.entry_points]
console_scripts =
dt-run = dt.main:main
[options.packages.find]
where = src
exclude =
tests*
# Settings for Flake8: Tool For Style Guide Enforcement
[flake8]
max-line-length = 120
exclude = venv/*
# Ignore completely:
# E203 - White space before ':', (conflicts with black)
# E231 - Missing whitespace after ',', ';', or ':'
# E731 - do not assign a lambda expression, use a def
# W503 - line break before binary operator, (conflicts with black)
# W605 - invalid escape sequence '\', (causes failures)
ignore = E203,E231,E731,W503,W605
# Settings for Mypy: static type checker for Python 3
[mypy]
ignore_missing_imports = True
[tool:pytest]
addopts =
# By default, we don't test models because doing so will
# make real requests and spend real money
-m 'not models'
markers =
# Marker for tests that make real model requests
models