Repository navigation
Expand file tree
/
Copy pathenv_base.yml
More file actions
40 lines (40 loc) · 1.52 KB
/
Copy pathenv_base.yml
File metadata and controls
40 lines (40 loc) · 1.52 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
channels:
- astra-toolbox
- conda-forge
- nvidia
dependencies:
- astra-toolbox>=2.3.1 # official `astra-toolbox` channel doesn't ship any `aarch64` build
# and `conda-forge` only ships astra-toolbox up to 2.3.1
- cuda-version=12.9 # conda-forge's CUDA builds of pytorch: 12.9 has torch 2.13, 12.8 only 2.7
- pip<=25.2 # pip 25.3 drops support for setup.py install which is used to install tomosipo and ts_algorithms from source
- python=3.12
# torch from conda too, so every compiled library uses conda's C++ runtime. A pip
# torch loads the system's, and conda's scipy etc. then fail on machines with an
# older system runtime ("CXXABI_1.3.15 not found").
# >= 2.6: torch.load(weights_only=True), with which LION loads model files, had a
# code-execution bug before 2.6 (CVE-2025-32434).
- pytorch-gpu>=2.6
# CUDA's headers (cuda.h): Triton, which comes with pytorch-gpu, compiles a small
# helper with them the first time it runs a kernel (LION's MSDNet on GPUs)
- cuda-cudart-dev
# LION does not import torchvision, but mrpro and spyrit (pip, pyproject.toml)
# require it: from conda here, or pip would install it with a pip torch
- torchvision
- deepinv # example scripts
- torchdiffeq
- torchmetrics # example scripts
- imageio
- matplotlib
- numpy<2.5 # numpy 2.5 (and above) breaks spgl1 0.0.3 (currently latest version)
- opencv
- pandas
- patchelf
- pillow
- pyside6
- pywavelets
- scikit-image
- scipy
- spgl1
- statsmodels
- tifffile
- typing_extensions>=4.16