-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathrequirements.txt
More file actions
80 lines (71 loc) · 2.79 KB
/
Copy pathrequirements.txt
File metadata and controls
80 lines (71 loc) · 2.79 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
-e / #tam debugging mode
#tests
pytest
pytest-cov
# Documentation
sphinx>=7.0.0
sphinx-rtd-theme
sphinxcontrib-bibtex
myst-parser
furo
# Data & scientific calculation
pandas>=2.0.0
numpy>=1.24.0
scipy>=1.10.0
# Visualisation
geopandas==1.1.1
matplotlib>=3.7.0
seaborn>=0.12.0
folium
ipython>=9.12.0
# ML & Stats
statsmodels>=0.14.0
scikit-learn>=1.3.0
pygam>=0.9.0
tabicl==2.1.1
tabicl[regressor]==2.1.1
tabicl[forecast]==2.1.1
## ==============================================================================
## TAM Hardware Configurations
## ==============================================================================
## The standard `pip install .` command uses PyPI defaults, which often
## downloads the massive CUDA (GPU) version of PyTorch on Windows/Linux.
## Use the options below to explicitly control your hardware setup.
##
## INSTRUCTIONS:
## To activate an option, remove the single '#' at the start of the relevant lines.
## Do not remove the double '##' (these are purely informational comments).
## --- Option 1: Force Lightweight CPU Version (Fast Loading) ---
## This forces the installation of the standard TAM CPU-only configuration.
## The framework loading time (`import tam as ta`) with this version is very fast: ~4 seconds.
##
## HOW TO SWITCH BACK FROM GPU TO CPU:
## 1. First, you must completely remove the GPU wheels by running this command:
## pip uninstall torch torchvision torchaudio
## 2. Uncomment the two lines below (remove the single '#').
## 3. Run: pip install -r requirements.txt
##
# --extra-index-url https://download.pytorch.org/whl/cpu
# torch>=2.4.1
## --- Option 2: Force Custom GPU Version (Local Wheels) ---
## Uncomment the lines below (remove the single '#') to install from your local
## files if you need hardware acceleration.
##
## HOW TO FIND THE RIGHT WHEELS FOR YOUR GPU:
## 1. Open your terminal/command prompt and type `nvidia-smi` to find your max supported CUDA version.
## 2. Go to the PyTorch wheel archive: https://download.pytorch.org/whl/torch_stable.html
## 3. Search for the files that match your specific setup:
## - The CUDA version (e.g., 'cu121' for CUDA 12.1).
## - Your Python version (e.g., 'cp312' for Python 3.12).
## - Your OS (e.g., 'win_amd64' for Windows).
## 4. Download the files, place them in the 'package/' folder, and update the filenames below if needed.
##
## PERFORMANCE WARNING FOR WINDOWS:
## Installing these massive CUDA wheels makes `import tam as ta` very slow:
## ~ 40 seconds initial load time when using the GPU.
## ~ 20 seconds initial load time even if you force CPU usage in your script via:
## os.environ["CUDA_VISIBLE_DEVICES"] = "-1"
##
# package/torch-2.5.1+cu121-cp312-cp312-win_amd64.whl
# package/torchvision-0.20.1+cu121-cp312-cp312-win_amd64.whl
# package/torchaudio-2.5.1+cu121-cp312-cp312-win_amd64.whl