-
Notifications
You must be signed in to change notification settings - Fork 4
Expand file tree
/
Copy pathrequirements.in
More file actions
92 lines (89 loc) · 3.14 KB
/
Copy pathrequirements.in
File metadata and controls
92 lines (89 loc) · 3.14 KB
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
# Keep always these for CBS
jupyterlab # notebook IDE
notebook # classic Jupyter notebook interface
ipython # interactive Python shell
jupyter # Jupyter meta-package
openpyxl # read/write Excel files
pyreadstat # read SPSS/SAS/Stata files
pandas # dataframes
numpy # arrays and numerical computing
scipy # scientific computing
polars # fast dataframes
duckdb # in-process SQL database
duckdb-engine # SQLAlchemy driver for duckdb
ibis-framework # dataframe API over SQL backends
matplotlib # plotting
seaborn # statistical plotting
networkx # graph analysis
igraph # fast graph analysis
netcbs # CBS-specific helpers
scikit-learn # classic ML
metasyn # synthetic data generation
metasyn-disclosure # disclosure control for synthetic data
PyYAML # YAML parsing
## Optional packages, specific for 9424
# Data handling
dask # parallel/out-of-core dataframes
h5py # HDF5 file format
numba # JIT compilation for numerical code
jax # array computing with autodiff/JIT
pyarrow # Arrow/Parquet file format
jupysql # run SQL in notebooks
# Visualization
plotly # interactive plots
hvplot # high-level plotting on top of holoviews
kaleido # static image export for plotly
# ML and interpretability
#catboost
#lime
eli5 # model inspection/explanation
shap # SHAP value model explanations
captum # PyTorch model interpretability
#imbalanced-learn
#interpret
optuna # hyperparameter optimization
#ray[tune]
# Strongly recommend the nightly build on torchtune so we can use all the custom functionality we've been building
# pip install --pre torchtune --extra-index-url https://download.pytorch.org/whl/nightly/cpu
#torchtune
#peft
hydra-core # config management
#hydra-optuna-sweeper
jupyterlab-optuna # Optuna dashboard in JupyterLab
# Statistics
statsmodels # statistical models
#pymc3 (pymc3 is at the current version not playing nicely with the other packages in pip)
lifelines # survival analysis
# Deep learning
torch==2.12.* # deep learning framework; must match the torch version in the --find-links URL below, or PyG's compiled wheels (pyg-lib, torch-sparse, etc.) will fail to load
#tensorflow (nit does not play nicely with pip)
torch-geometric # graph neural networks
# PyG CPU wheels are not on the normal package index, only at this URL:
--find-links https://data.pyg.org/whl/torch-2.12.0+cpu.html
pyg-lib # compiled ops for PyG
torch-sparse # sparse tensor ops for PyG
torch-scatter # scatter ops for PyG
#torch-cluster # clustering ops for PyG deprecated
#torch-spline-conv # spline-based conv ops for PyG
#torch-quiver #not yet available for windows
#intel-extension-for-pytorch #not yet available for windows
#keras
#dgl
gensim # topic modeling/word embeddings
#pytorch-lightning
transformers # pretrained NLP/LLM models
umap-learn # dimensionality reduction
#inspect-ai
torchao # PyTorch quantization/optimization
wandb # experiment tracking
#mlflow
tensorboard # training visualization
flash-attn; sys_platform == "linux" #attention-based learning faster; CUDA build, no official wheels for macOS/Windows
accelerate # multi-GPU/mixed-precision training helper
#annoy
# Configuration files
json5 # JSON5 parsing
## Testing
pytest # testing framework
ruff # linter/formatter
pre-commit # git hook manager