search

openmm / openmm-ml

High level API for using machine learning models in OpenMM simulations
Other
75 stars 25 forks source link

Failure to import openmmtorch #67

Closed rafwiewiora closed 5 months ago

rafwiewiora commented 5 months ago

I'm stuck trying to make dependencies work, related to https://github.com/openmm/openmm-torch/issues/127.

After standard mamba installation of openmm-ml I get: 

File ~/miniforge3/envs/openmm-ml/lib/python3.11/site-packages/openmmml/models/anipotential.py:80, in ANIPotentialImpl.addForces(self, topology, system, atoms, forceGroup, filename, implementation, modelIndex, **args)
     78 import torchani
     79 import torch
---> 80 import openmmtorch
     82 # `nnpops` throws error if `periodic_table_index`=False if one passes `species` as `species_to_tensor` from `element`
     83 _kwarg_dict = {'periodic_table_index': True}

File ~/miniforge3/envs/openmm-ml/lib/python3.11/site-packages/openmmtorch.py:15
     13     from . import _openmmtorch
     14 else:
---> 15     import _openmmtorch
     17 try:
     18     import builtins as __builtin__

ImportError: libtorch_cuda.so: cannot open shared object file: No such file or directory

on trying to createSystem. I have the same problem with separately installing openmmtorch and this solution worked. But trying to install openmm-ml on top of that working openmmtorch (which requires using --no-deps) leads to the 'undefined symbol' error from torchani import, as seen here.

peastman commented 5 months ago

Can you post conda list for your environment? This probably means you have the CPU version of PyTorch installed.

We've been running into a lot of conda issues lately with PyTorch, CUDA, OpenMM, and other packages. See for example https://github.com/torchmd/torchmd-net/issues/266 and https://github.com/conda-forge/openmm-torch-feedstock/issues/48. I'm not sure what the cause is. @RaulPPelaez might have some insight.

RaulPPelaez commented 5 months ago

I do not know what is going on these days. Seems like conda is falling apart -.-

Please provide the commands you ran from a clean environment and the result of conda-list so we can provide more help.

Be careful with settings like CONDA_CUDA_OVERRIDE, they are tricky and may cause unexpected inconsistencies.

For me, running something like this:

mamba create -n test openmm-torch "cuda-version>=12" openmm-ml

Tries to install the cpu versions of pytorch and some other packages, which is incorrect behavior. I also checked with conda and same behavior. I have to specifically ask for pytorch cuda like this:

mamba create -n test openmm-torch openmm-ml pytorch=*=*cuda*

This results in an env in which I can import openmm-ml without issue.

There are these commented lines in the openmm-torch feedstock:

  #run_constrained:
    # 2022/02/05 hmaarrfk
    # While conda packaging seems to allow us to specify
    # constraints on the same package in different lines
    # the resulting package doesn't have the ability to
    # be specified in multiples lines
    # This makes it tricky to use run_exports
    # we add the GPU constraint in the run_constrained
    # to allow us to have "two" constraints on the
    # running package
    #- pytorch =*={{ torch_proc_type }}*

It really looks to me like that should be uncommented. Lets try...

RaulPPelaez commented 5 months ago

Working on it here https://github.com/conda-forge/openmm-torch-feedstock/pull/49

rafwiewiora commented 5 months ago

Thank you both! mamba install openmm-torch openmm-ml pytorch=*=*cuda* indeed did the magic. For completeness, here's the conda list I get after mamba install openmm-ml -- it is the CPU version of pytorch problem:

# packages in environment at /home/rafal.wiewiora/miniforge3/envs/openmmmlbroken:
#
# Name                    Version                   Build  Channel
_libgcc_mutex             0.1                 conda_forge    conda-forge
_openmp_mutex             4.5                       2_gnu    conda-forge
brotli-python             1.1.0           py311hb755f60_1    conda-forge
bzip2                     1.0.8                hd590300_5    conda-forge
c-ares                    1.26.0               hd590300_0    conda-forge
ca-certificates           2024.2.2             hbcca054_0    conda-forge
cached-property           1.5.2                hd8ed1ab_1    conda-forge
cached_property           1.5.2              pyha770c72_1    conda-forge
certifi                   2023.11.17         pyhd8ed1ab_0    conda-forge
charset-normalizer        3.3.2              pyhd8ed1ab_0    conda-forge
cudatoolkit               11.8.0              h4ba93d1_13    conda-forge
filelock                  3.13.1             pyhd8ed1ab_0    conda-forge
fsspec                    2023.12.2          pyhca7485f_0    conda-forge
gmp                       6.3.0                h59595ed_0    conda-forge
gmpy2                     2.1.2           py311h6a5fa03_1    conda-forge
h5py                      3.10.0          nompi_py311hebc2b07_101    conda-forge
hdf5                      1.14.3          nompi_h4f84152_100    conda-forge
idna                      3.6                pyhd8ed1ab_0    conda-forge
importlib-metadata        7.0.1              pyha770c72_0    conda-forge
importlib_metadata        7.0.1                hd8ed1ab_0    conda-forge
jinja2                    3.1.3              pyhd8ed1ab_0    conda-forge
keyutils                  1.6.1                h166bdaf_0    conda-forge
krb5                      1.21.2               h659d440_0    conda-forge
lark-parser               0.12.0             pyhd8ed1ab_0    conda-forge
ld_impl_linux-64          2.40                 h41732ed_0    conda-forge
libabseil                 20230802.1      cxx17_h59595ed_0    conda-forge
libaec                    1.1.2                h59595ed_1    conda-forge
libblas                   3.9.0           21_linux64_openblas    conda-forge
libcblas                  3.9.0           21_linux64_openblas    conda-forge
libcurl                   8.5.0                hca28451_0    conda-forge
libedit                   3.1.20191231         he28a2e2_2    conda-forge
libev                     4.33                 hd590300_2    conda-forge
libexpat                  2.5.0                hcb278e6_1    conda-forge
libffi                    3.4.2                h7f98852_5    conda-forge
libgcc-ng                 13.2.0               h807b86a_5    conda-forge
libgfortran-ng            13.2.0               h69a702a_5    conda-forge
libgfortran5              13.2.0               ha4646dd_5    conda-forge
libgomp                   13.2.0               h807b86a_5    conda-forge
liblapack                 3.9.0           21_linux64_openblas    conda-forge
libnghttp2                1.58.0               h47da74e_1    conda-forge
libnsl                    2.0.1                hd590300_0    conda-forge
libopenblas               0.3.26          pthreads_h413a1c8_0    conda-forge
libprotobuf               4.25.1               hf27288f_0    conda-forge
libsqlite                 3.44.2               h2797004_0    conda-forge
libssh2                   1.11.0               h0841786_0    conda-forge
libstdcxx-ng              13.2.0               h7e041cc_5    conda-forge
libtorch                  2.1.2           cpu_generic_ha017de0_1    conda-forge
libuuid                   2.38.1               h0b41bf4_0    conda-forge
libuv                     1.46.0               hd590300_0    conda-forge
libxcrypt                 4.4.36               hd590300_1    conda-forge
libzlib                   1.2.13               hd590300_5    conda-forge
markupsafe                2.1.4           py311h459d7ec_0    conda-forge
mpc                       1.3.1                hfe3b2da_0    conda-forge
mpfr                      4.2.1                h9458935_0    conda-forge
mpmath                    1.3.0              pyhd8ed1ab_0    conda-forge
ncurses                   6.4                  h59595ed_2    conda-forge
networkx                  3.2.1              pyhd8ed1ab_0    conda-forge
nomkl                     1.0                  h5ca1d4c_0    conda-forge
numpy                     1.26.3          py311h64a7726_0    conda-forge
ocl-icd                   2.3.1                h7f98852_0    conda-forge
ocl-icd-system            1.0.0                         1    conda-forge
openmm                    8.1.1           py311h9766050_0    conda-forge
openmm-ml                 1.1                pyhd8ed1ab_0    conda-forge
openmm-torch              1.4             cpu_py311h446247e_3    conda-forge
openssl                   3.2.1                hd590300_0    conda-forge
packaging                 23.2               pyhd8ed1ab_0    conda-forge
pip                       23.3.2             pyhd8ed1ab_0    conda-forge
pysocks                   1.7.1              pyha2e5f31_6    conda-forge
python                    3.11.7          hab00c5b_1_cpython    conda-forge
python_abi                3.11                    4_cp311    conda-forge
pytorch                   2.1.2           cpu_generic_py311h1584bb0_1    conda-forge
readline                  8.2                  h8228510_1    conda-forge
requests                  2.31.0             pyhd8ed1ab_0    conda-forge
scipy                     1.12.0          py311h64a7726_2    conda-forge
setuptools                65.3.0             pyhd8ed1ab_1    conda-forge
setuptools-scm            6.3.2              pyhd8ed1ab_0    conda-forge
setuptools_scm            6.3.2                hd8ed1ab_0    conda-forge
sleef                     3.5.1                h9b69904_2    conda-forge
sympy                     1.12            pypyh9d50eac_103    conda-forge
tk                        8.6.13          noxft_h4845f30_101    conda-forge
tomli                     2.0.1              pyhd8ed1ab_0    conda-forge
torchani                  2.2.4           cpu_py311h12a0d1d_3    conda-forge
typing_extensions         4.9.0              pyha770c72_0    conda-forge
tzdata                    2024a                h0c530f3_0    conda-forge
urllib3                   2.2.0              pyhd8ed1ab_0    conda-forge
wheel                     0.42.0             pyhd8ed1ab_0    conda-forge
xz                        5.2.6                h166bdaf_0    conda-forge
zipp                      3.17.0             pyhd8ed1ab_0    conda-forge
zstd                      1.5.5                hfc55251_0    conda-forge
peastman commented 5 months ago

We have new packages for OpenMM and OpenMM-Torch. Hopefully they fix the problem. Can you try again and see?

rafwiewiora commented 5 months ago

Confirmed, no problems now, thanks again!