deepmodeling / abacus-develop

An electronic structure package based on either plane wave basis or numerical atomic orbitals.
http://abacus.ustc.edu.cn
GNU Lesser General Public License v3.0
174 stars 136 forks source link

Questions about the use of DeePKS #5434

Closed yycx1111 closed 1 week ago

yycx1111 commented 2 weeks ago

Details

I have installed deepks and abacus on my local machine. I met some problems with package and function versions in the case of running a single water molecule in deepks. Here are a few of my questions, since I didn't get feedback for asking questions in the deepks community. I ask the same questions in abacus.

  1. What version of pyscf is compatible with deepks? What can be done about the following problems? /deepks-kit/examples/water_single/iter/iter.init/00.scf/task.trn.00/err: Traceback (most recent call last): File "", line 198, in _run_module_as_main File "", line 88, in _run_code File "/home/ps/miniconda3/envs/deepks/lib/python3.12/site-packages/deepks/scf/run.py", line 6, in from pyscf import gto, lib ModuleNotFoundError: No module named 'pyscf' Traceback (most recent call last): File "", line 198, in _run_module_as_main File "", line 88, in _run_code File "/home/ps/miniconda3/envs/deepks/lib/python3.12/site-packages/deepks/scf/run.py", line 6, in from pyscf import gto, lib ModuleNotFoundError: No module named 'pyscf' Traceback (most recent call last): File "", line 198, in _run_module_as_main File "", line 88, in _run_code File "/home/ps/miniconda3/envs/deepks/lib/python3.12/site-packages/deepks/scf/run.py", line 6, in from pyscf import gto, lib ModuleNotFoundError: No module named 'pyscf' Traceback (most recent call last): File "", line 198, in _run_module_as_main File "", line 88, in _run_code File "/home/ps/miniconda3/envs/deepks/lib/python3.12/site-packages/deepks/scf/run.py", line 6, in from pyscf import gto, lib ModuleNotFoundError: No module named 'pyscf' systems/group.00 failed! error: PenaltyMixin.get_fock() got an unexpected keyword argument 'fock_last' Traceback (most recent call last): File "", line 198, in _run_module_as_main File "", line 88, in _run_code File "/home/ps/miniconda3/envs/deepks/lib/python3.12/site-packages/deepks/scf/run.py", line 268, in cli() File "/home/ps/miniconda3/envs/deepks/lib/python3.12/site-packages/deepks/main.py", line 174, in scf_cli main(argdict) File "/home/ps/miniconda3/envs/deepks/lib/python3.12/site-packages/deepks/scf/run.py", line 236, in main meta, result = solve_mol(mol, model, fields, labels, ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/home/ps/miniconda3/envs/deepks/lib/python3.12/site-packages/deepks/scf/run.py", line 52, in solve_mol cf.kernel() File "", line 2, in kernel File "/home/ps/miniconda3/envs/deepks/lib/python3.12/site-packages/pyscf/scf/hf.py", line 2000, in scf kernel(self, self.conv_tol, self.conv_tol_grad, File "/home/ps/miniconda3/envs/deepks/lib/python3.12/site-packages/pyscf/scf/hf.py", line 175, in kernel fock = mf.get_fock(h1e, s1e, vhf, dm, cycle, mf_diis, fock_last=fock_last) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ TypeError: PenaltyMixin.get_fock() got an unexpected keyword argument 'fock_last' systems/group.00 failed! error: PenaltyMixin.get_fock() got an unexpected keyword argument 'fock_last' Traceback (most recent call last): File "", line 198, in _run_module_as_main File "", line 88, in _run_code File "/home/ps/miniconda3/envs/deepks/lib/python3.12/site-packages/deepks/scf/run.py", line 268, in cli() File "/home/ps/miniconda3/envs/deepks/lib/python3.12/site-packages/deepks/main.py", line 174, in scf_cli main(argdict) File "/home/ps/miniconda3/envs/deepks/lib/python3.12/site-packages/deepks/scf/run.py", line 236, in main meta, result = solve_mol(mol, model, fields, labels, ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/home/ps/miniconda3/envs/deepks/lib/python3.12/site-packages/deepks/scf/run.py", line 52, in solve_mol cf.kernel() File "", line 2, in kernel File "/home/ps/miniconda3/envs/deepks/lib/python3.12/site-packages/pyscf/scf/hf.py", line 2000, in scf kernel(self, self.conv_tol, self.conv_tol_grad, File "/home/ps/miniconda3/envs/deepks/lib/python3.12/site-packages/pyscf/scf/hf.py", line 175, in kernel fock = mf.get_fock(h1e, s1e, vhf, dm, cycle, mf_diis, fock_last=fock_last) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ TypeError: PenaltyMixin.get_fock() got an unexpected keyword argument 'fock_last' systems/group.00 failed! error: PenaltyMixin.get_fock() got an unexpected keyword argument 'fock_last' Traceback (most recent call last): File "", line 198, in _run_module_as_main File "", line 88, in _run_code File "/home/ps/miniconda3/envs/deepks/lib/python3.12/site-packages/deepks/scf/run.py", line 268, in cli() File "/home/ps/miniconda3/envs/deepks/lib/python3.12/site-packages/deepks/main.py", line 174, in scf_cli main(**argdict) File "/home/ps/miniconda3/envs/deepks/lib/python3.12/site-packages/deepks/scf/run.py", line 236, in main meta, result = solve_mol(mol, model, fields, labels, ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/home/ps/miniconda3/envs/deepks/lib/python3.12/site-packages/deepks/scf/run.py", line 52, in solve_mol cf.kernel() File "", line 2, in kernel File "/home/ps/miniconda3/envs/deepks/lib/python3.12/site-packages/pyscf/scf/hf.py", line 2000, in scf kernel(self, self.conv_tol, self.conv_tol_grad, File "/home/ps/miniconda3/envs/deepks/lib/python3.12/site-packages/pyscf/scf/hf.py", line 175, in kernel fock = mf.get_fock(h1e, s1e, vhf, dm, cycle, mf_diis, fock_last=fock_last) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ TypeError: PenaltyMixin.get_fock() got an unexpected keyword argument 'fock_last'

  2. Is deepks still being updated? Cause I've encountered some problems that suggest that the function or package used is old and may no longer work.

  3. Is there a container image with deepks readily available on the Bohrium platform?

Have you read FAQ on the online manual http://abacus.deepmodeling.com/en/latest/community/faq.html

Task list for Issue attackers (only for developers)

QuantumMisaka commented 2 weeks ago

@WHUweiqingzhou Any comments? I do think a hands-on tutorial for deepks-abacus installation along with deepks notebook is needed

WHUweiqingzhou commented 2 weeks ago

@WHUweiqingzhou Any comments? I do think a hands-on tutorial for deepks-abacus installation along with deepks notebook is needed

Second it. We will consider this.

ErjieWu commented 1 week ago

@yycx1111 It seems there are issues with matching deepks-kit with the newest pyscf version. You can use deepks-kit with ABACUS instead.

Here's a way to use ABACUS+deepks in conda environment:

conda create -n abacus+deepks python=3.12 numpy=1.26.0 scipy h5py ruamel.yaml=0.17.21 paramiko abacus "libblas=*=*mkl" mpich -c conda-forge

conda activate abacus+deepks

# Install pytorch, the cuda version depends on your machine
conda install pytorch torchvision torchaudio cpuonly -c pytorch # cpu version
conda install pytorch torchvision torchaudio pytorch-cuda=12.4 -c pytorch -c nvidia # gpu version

# use develop branch
pip install git+https://github.com/deepmodeling/deepks-kit@develop
yycx1111 commented 1 week ago

@yycx1111 It seems there are issues with matching deepks-kit with the newest pyscf version. You can use deepks-kit with ABACUS instead.

Here's a way to use ABACUS+deepks in conda environment:

conda create -n abacus+deepks python=3.12 numpy=1.26.0 scipy h5py ruamel.yaml=0.17.21 paramiko abacus "libblas=*=*mkl" mpich -c conda-forge

conda  activate abacus+deepks

# Install pytorch, the cuda version depends on your machine
conda install pytorch cudatoolkit=10.2 -c pytorch # cpu version
conda install pytorch torchvision torchaudio pytorch-cuda=12.4 -c pytorch -c nvidia # gpu version

# use develop branch
pip install git+https://github.com/deepmodeling/deepks-kit@develop

Thank you for your reply. But I have been experiencing problems with network connections or failing to connect to the site when creating the environment, and I still have problems with connection errors even after switching to the Tsinghua source. Do you have any suggestions on how to create an environment? The problems are like this: CondaHTTPError: HTTP 000 CONNECTION FAILED for url http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge/linux-64/libmagma-2.8.0-hfdb99dd_0.conda
Elapsed: -
An HTTP error occurred when trying to retrieve this URL.
HTTP errors are often intermittent, and a simple retry will get you on your way.
CondaHTTPError: HTTP 000 CONNECTION FAILED for url http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge/linux-64/libmagma-2.8.0-hfdb99dd_0.conda
Elapsed: -
An HTTP error occurred when trying to retrieve this URL.
HTTP errors are often intermittent, and a simple retry will get you on your way.

ErjieWu commented 1 week ago

@yycx1111 It seems there are issues with matching deepks-kit with the newest pyscf version. You can use deepks-kit with ABACUS instead. Here's a way to use ABACUS+deepks in conda environment:

conda create -n abacus+deepks python=3.12 numpy=1.26.0 scipy h5py ruamel.yaml=0.17.21 paramiko abacus "libblas=*=*mkl" mpich -c conda-forge

conda  activate abacus+deepks

# Install pytorch, the cuda version depends on your machine
conda install pytorch cudatoolkit=10.2 -c pytorch # cpu version
conda install pytorch torchvision torchaudio pytorch-cuda=12.4 -c pytorch -c nvidia # gpu version

# use develop branch
pip install git+https://github.com/deepmodeling/deepks-kit@develop

Thank you for your reply. But I have been experiencing problems with network connections or failing to connect to the site when creating the environment, and I still have problems with connection errors even after switching to the Tsinghua source. Do you have any suggestions on how to create an environment? The problems are like this: CondaHTTPError: HTTP 000 CONNECTION FAILED for url http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge/linux-64/libmagma-2.8.0-hfdb99dd_0.conda Elapsed: - An HTTP error occurred when trying to retrieve this URL. HTTP errors are often intermittent, and a simple retry will get you on your way. CondaHTTPError: HTTP 000 CONNECTION FAILED for url http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge/linux-64/libmagma-2.8.0-hfdb99dd_0.conda Elapsed: - An HTTP error occurred when trying to retrieve this URL. HTTP errors are often intermittent, and a simple retry will get you on your way.

This looks like a problem with your network setup. Maybe you should check things like your network environment and proxy settings.

yycx1111 commented 1 week ago

@yycx1111 It seems there are issues with matching deepks-kit with the newest pyscf version. You can use deepks-kit with ABACUS instead. Here's a way to use ABACUS+deepks in conda environment:

conda create -n abacus+deepks python=3.12 numpy=1.26.0 scipy h5py ruamel.yaml=0.17.21 paramiko abacus "libblas=*=*mkl" mpich -c conda-forge

conda  activate abacus+deepks

# Install pytorch, the cuda version depends on your machine
conda install pytorch cudatoolkit=10.2 -c pytorch # cpu version
conda install pytorch torchvision torchaudio pytorch-cuda=12.4 -c pytorch -c nvidia # gpu version

# use develop branch
pip install git+https://github.com/deepmodeling/deepks-kit@develop

Thank you for your reply. But I have been experiencing problems with network connections or failing to connect to the site when creating the environment, and I still have problems with connection errors even after switching to the Tsinghua source. Do you have any suggestions on how to create an environment? The problems are like this: CondaHTTPError: HTTP 000 CONNECTION FAILED for url http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge/linux-64/libmagma-2.8.0-hfdb99dd_0.conda Elapsed: - An HTTP error occurred when trying to retrieve this URL. HTTP errors are often intermittent, and a simple retry will get you on your way. CondaHTTPError: HTTP 000 CONNECTION FAILED for url http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge/linux-64/libmagma-2.8.0-hfdb99dd_0.conda Elapsed: - An HTTP error occurred when trying to retrieve this URL. HTTP errors are often intermittent, and a simple retry will get you on your way.

This looks like a problem with your network setup. Maybe you should check things like your network environment and proxy settings.

After running conda install pytorch cudatoolkit=10.2 -c pytorch # cpu version, I encountered the following problems. Any suggestions for the pytorch install command for CPU version? Channels:

LibMambaUnsatisfiableError: Encountered problems while solving:

Could not solve for environment specs
The following packages are incompatible
├─ abacus is installable and it requires
│ └─ libtorch >=2.5.1,<2.6.0a0 , which requires │ ├─ nccl >=2.23.4.1,<3.0a0 , which requires │ │ └─ cuda-version >=11.0,<12.0a0 with the potential options │ │ ├─ cuda-version 11.8 would require │ │ │ └─ cudatoolkit 11.8|11.8. , which can be installed; │ │ ├─ cuda-version 11.0 would require │ │ │ └─ cudatoolkit 11.0|11.0. , which can be installed; │ │ ├─ cuda-version 11.1 would require │ │ │ └─ cudatoolkit 11.1|11.1. , which can be installed; │ │ ├─ cuda-version 11.2 would require │ │ │ └─ cudatoolkit 11.2|11.2. , which can be installed; │ │ ├─ cuda-version 11.3 would require │ │ │ └─ cudatoolkit 11.3|11.3. , which can be installed; │ │ ├─ cuda-version 11.4 would require │ │ │ └─ cudatoolkit 11.4|11.4. , which can be installed; │ │ ├─ cuda-version 11.5 would require │ │ │ └─ cudatoolkit 11.5|11.5. , which can be installed; │ │ ├─ cuda-version 11.6 would require │ │ │ └─ cudatoolkit 11.6|11.6. , which can be installed; │ │ └─ cuda-version 11.7 would require │ │ └─ cudatoolkit 11.7|11.7. , which can be installed; │ └─ pytorch 2.5.1 cuda118__303, which can be installed; ├─ cudatoolkit 10.2* is not installable because it conflicts with any installable versions previously reported; └─ pytorch is not installable because there are no viable options ├─ pytorch [0.4.0|0.4.1|...|2.5.1] conflicts with any installable versions previously reported; ├─ pytorch [1.10.0|1.10.1|...|1.9.1] would require │ └─ cudatoolkit >=11.1,<11.2 , which does not exist (perhaps a missing channel); ├─ pytorch 1.11.0 would require │ └─ cudatoolkit >=11.5,<11.6 , which does not exist (perhaps a missing channel); ├─ pytorch [1.12.0|1.12.1] would require │ └─ cudatoolkit >=11.6,<11.7 , which does not exist (perhaps a missing channel); ├─ pytorch [1.13.0|1.13.1] would require │ └─ pytorch-cuda >=11.6,<11.7 , which requires │ └─ cuda 11.6. , which does not exist (perhaps a missing channel); ├─ pytorch [1.13.0|1.13.1|2.0.0|2.0.1] would require │ └─ pytorch-cuda >=11.7,<11.8 but there are no viable options │ ├─ pytorch-cuda 11.7 would require │ │ └─ cuda-cudart >=11.7,<11.8 , which does not exist (perhaps a missing channel); │ └─ pytorch-cuda 11.7 would require │ └─ cuda 11.7. , which does not exist (perhaps a missing channel); ├─ pytorch [1.7.1|1.8.0] would require │ └─ python_abi 3.9. _cp39, which does not exist (perhaps a missing channel); ├─ pytorch [2.0.0|2.0.1|...|2.5.1] would require │ └─ pytorch-cuda >=11.8,<11.9 but there are no viable options │ ├─ pytorch-cuda 11.8 would require │ │ └─ cuda-cudart >=11.8,<12.0 , which does not exist (perhaps a missing channel); │ └─ pytorch-cuda 11.8 would require │ └─ cuda 11.8. , which does not exist (perhaps a missing channel); ├─ pytorch [2.1.0|2.1.1|...|2.5.1] would require │ └─ pytorch-cuda >=12.1,<12.2 , which requires │ └─ cuda-cudart >=12.1,<12.2 , which does not exist (perhaps a missing channel); ├─ pytorch [0.2.0|0.3.0] would require │ └─ cudatoolkit 7.5 , which does not exist (perhaps a missing channel); ├─ pytorch [0.2.0|0.3.0|0.3.1] would require │ └─ cudatoolkit 8.0 , which does not exist (perhaps a missing channel); ├─ pytorch [0.3.1|0.4.0] would require │ └─ cudatoolkit 8.0.* , which does not exist (perhaps a missing channel); ├─ pytorch 1.0.1 would require │ └─ cudatoolkit >=8.0,<8.1.0a0 , which does not exist (perhaps a missing channel); └─ pytorch [2.0.1|2.3.0] would require └─ __cuda >=11.8 , which is missing on the system.

ErjieWu commented 1 week ago

@yycx1111 It seems there are issues with matching deepks-kit with the newest pyscf version. You can use deepks-kit with ABACUS instead. Here's a way to use ABACUS+deepks in conda environment:

conda create -n abacus+deepks python=3.12 numpy=1.26.0 scipy h5py ruamel.yaml=0.17.21 paramiko abacus "libblas=*=*mkl" mpich -c conda-forge

conda  activate abacus+deepks

# Install pytorch, the cuda version depends on your machine
conda install pytorch cudatoolkit=10.2 -c pytorch # cpu version
conda install pytorch torchvision torchaudio pytorch-cuda=12.4 -c pytorch -c nvidia # gpu version

# use develop branch
pip install git+https://github.com/deepmodeling/deepks-kit@develop

Thank you for your reply. But I have been experiencing problems with network connections or failing to connect to the site when creating the environment, and I still have problems with connection errors even after switching to the Tsinghua source. Do you have any suggestions on how to create an environment? The problems are like this: CondaHTTPError: HTTP 000 CONNECTION FAILED for url http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge/linux-64/libmagma-2.8.0-hfdb99dd_0.conda Elapsed: - An HTTP error occurred when trying to retrieve this URL. HTTP errors are often intermittent, and a simple retry will get you on your way. CondaHTTPError: HTTP 000 CONNECTION FAILED for url http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge/linux-64/libmagma-2.8.0-hfdb99dd_0.conda Elapsed: - An HTTP error occurred when trying to retrieve this URL. HTTP errors are often intermittent, and a simple retry will get you on your way.

This looks like a problem with your network setup. Maybe you should check things like your network environment and proxy settings.

After running conda install pytorch cudatoolkit=10.2 -c pytorch # cpu version, I encountered the following problems. Any suggestions for the pytorch install command for CPU version? Channels:

  • pytorch
  • defaults Platform: linux-64 Collecting package metadata (repodata.json): done Solving environment: failed

LibMambaUnsatisfiableError: Encountered problems while solving:

  • nothing provides cudatoolkit >=11.1,<11.2 needed by pytorch-1.10.0-py3.6_cuda11.1_cudnn8.0.5_0
  • package abacus-3.8.2-cuda_mpi_mpich_h6026299_1 requires libtorch >=2.5.1,<2.6.0a0, but none of the providers can be installed

Could not solve for environment specs The following packages are incompatible ├─ abacus is installable and it requires │ └─ libtorch >=2.5.1,<2.6.0a0 , which requires │ ├─ nccl >=2.23.4.1,<3.0a0 , which requires │ │ └─ cuda-version >=11.0,<12.0a0 with the potential options │ │ ├─ cuda-version 11.8 would require │ │ │ └─ cudatoolkit 11.8|11.8. , which can be installed; │ │ ├─ cuda-version 11.0 would require │ │ │ └─ cudatoolkit 11.0|11.0. , which can be installed; │ │ ├─ cuda-version 11.1 would require │ │ │ └─ cudatoolkit 11.1|11.1. , which can be installed; │ │ ├─ cuda-version 11.2 would require │ │ │ └─ cudatoolkit 11.2|11.2. , which can be installed; │ │ ├─ cuda-version 11.3 would require │ │ │ └─ cudatoolkit 11.3|11.3. , which can be installed; │ │ ├─ cuda-version 11.4 would require │ │ │ └─ cudatoolkit 11.4|11.4. , which can be installed; │ │ ├─ cuda-version 11.5 would require │ │ │ └─ cudatoolkit 11.5|11.5. , which can be installed; │ │ ├─ cuda-version 11.6 would require │ │ │ └─ cudatoolkit 11.6|11.6. , which can be installed; │ │ └─ cuda-version 11.7 would require │ │ └─ cudatoolkit 11.7|11.7.* , which can be installed; │ └─ pytorch 2.5.1 cuda118303, which can be installed; ├─ cudatoolkit 10.2** is not installable because it conflicts with any installable versions previously reported; └─ pytorch is not installable because there are no viable options ├─ pytorch [0.4.0|0.4.1|...|2.5.1] conflicts with any installable versions previously reported; ├─ pytorch [1.10.0|1.10.1|...|1.9.1] would require │ └─ cudatoolkit >=11.1,<11.2 , which does not exist (perhaps a missing channel); ├─ pytorch 1.11.0 would require │ └─ cudatoolkit >=11.5,<11.6 , which does not exist (perhaps a missing channel); ├─ pytorch [1.12.0|1.12.1] would require │ └─ cudatoolkit >=11.6,<11.7 , which does not exist (perhaps a missing channel); ├─ pytorch [1.13.0|1.13.1] would require │ └─ pytorch-cuda >=11.6,<11.7 , which requires │ └─ cuda 11.6. , which does not exist (perhaps a missing channel); ├─ pytorch [1.13.0|1.13.1|2.0.0|2.0.1] would require │ └─ pytorch-cuda >=11.7,<11.8 but there are no viable options │ ├─ pytorch-cuda 11.7 would require │ │ └─ cuda-cudart >=11.7,<11.8 , which does not exist (perhaps a missing channel); │ └─ pytorch-cuda 11.7 would require │ └─ cuda 11.7. , which does not exist (perhaps a missing channel); ├─ pytorch [1.7.1|1.8.0] would require │ └─ python_abi 3.9. _cp39, which does not exist (perhaps a missing channel); ├─ pytorch [2.0.0|2.0.1|...|2.5.1] would require │ └─ pytorch-cuda >=11.8,<11.9 but there are no viable options │ ├─ pytorch-cuda 11.8 would require │ │ └─ cuda-cudart >=11.8,<12.0 , which does not exist (perhaps a missing channel); │ └─ pytorch-cuda 11.8 would require │ └─ cuda 11.8. , which does not exist (perhaps a missing channel); ├─ pytorch [2.1.0|2.1.1|...|2.5.1] would require │ └─ pytorch-cuda >=12.1,<12.2 , which requires │ └─ cuda-cudart >=12.1,<12.2 , which does not exist (perhaps a missing channel); ├─ pytorch [0.2.0|0.3.0] would require │ └─ cudatoolkit 7.5 , which does not exist (perhaps a missing channel); ├─ pytorch [0.2.0|0.3.0|0.3.1] would require │ └─ cudatoolkit 8.0 , which does not exist (perhaps a missing channel); ├─ pytorch [0.3.1|0.4.0] would require │ └─ cudatoolkit 8.0.* , which does not exist (perhaps a missing channel); ├─ pytorch 1.0.1 would require │ └─ cudatoolkit >=8.0,<8.1.0a0 , which does not exist (perhaps a missing channel); └─ pytorch [2.0.1|2.3.0] would require └─ cuda >=11.8 , which is missing on the system.

Ahh, I’ve made a mistake. For cpu version, try

conda install pytorch torchvision torchaudio cpuonly -c pytorch
yycx1111 commented 1 week ago

@yycx1111 It seems there are issues with matching deepks-kit with the newest pyscf version. You can use deepks-kit with ABACUS instead. Here's a way to use ABACUS+deepks in conda environment:

conda create -n abacus+deepks python=3.12 numpy=1.26.0 scipy h5py ruamel.yaml=0.17.21 paramiko abacus "libblas=*=*mkl" mpich -c conda-forge

conda  activate abacus+deepks

# Install pytorch, the cuda version depends on your machine
conda install pytorch cudatoolkit=10.2 -c pytorch # cpu version
conda install pytorch torchvision torchaudio pytorch-cuda=12.4 -c pytorch -c nvidia # gpu version

# use develop branch
pip install git+https://github.com/deepmodeling/deepks-kit@develop

Thank you for your reply. But I have been experiencing problems with network connections or failing to connect to the site when creating the environment, and I still have problems with connection errors even after switching to the Tsinghua source. Do you have any suggestions on how to create an environment? The problems are like this: CondaHTTPError: HTTP 000 CONNECTION FAILED for url http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge/linux-64/libmagma-2.8.0-hfdb99dd_0.conda Elapsed: - An HTTP error occurred when trying to retrieve this URL. HTTP errors are often intermittent, and a simple retry will get you on your way. CondaHTTPError: HTTP 000 CONNECTION FAILED for url http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge/linux-64/libmagma-2.8.0-hfdb99dd_0.conda Elapsed: - An HTTP error occurred when trying to retrieve this URL. HTTP errors are often intermittent, and a simple retry will get you on your way.

This looks like a problem with your network setup. Maybe you should check things like your network environment and proxy settings.

After running conda install pytorch cudatoolkit=10.2 -c pytorch # cpu version, I encountered the following problems. Any suggestions for the pytorch install command for CPU version? Channels:

  • pytorch
  • defaults Platform: linux-64 Collecting package metadata (repodata.json): done Solving environment: failed

LibMambaUnsatisfiableError: Encountered problems while solving:

  • nothing provides cudatoolkit >=11.1,<11.2 needed by pytorch-1.10.0-py3.6_cuda11.1_cudnn8.0.5_0
  • package abacus-3.8.2-cuda_mpi_mpich_h6026299_1 requires libtorch >=2.5.1,<2.6.0a0, but none of the providers can be installed

Could not solve for environment specs The following packages are incompatible ├─ abacus is installable and it requires │ └─ libtorch >=2.5.1,<2.6.0a0 , which requires │ ├─ nccl >=2.23.4.1,<3.0a0 , which requires │ │ └─ cuda-version >=11.0,<12.0a0 with the potential options │ │ ├─ cuda-version 11.8 would require │ │ │ └─ cudatoolkit 11.8|11.8. , which can be installed; │ │ ├─ cuda-version 11.0 would require │ │ │ └─ cudatoolkit 11.0|11.0. , which can be installed; │ │ ├─ cuda-version 11.1 would require │ │ │ └─ cudatoolkit 11.1|11.1. , which can be installed; │ │ ├─ cuda-version 11.2 would require │ │ │ └─ cudatoolkit 11.2|11.2. , which can be installed; │ │ ├─ cuda-version 11.3 would require │ │ │ └─ cudatoolkit 11.3|11.3. , which can be installed; │ │ ├─ cuda-version 11.4 would require │ │ │ └─ cudatoolkit 11.4|11.4. , which can be installed; │ │ ├─ cuda-version 11.5 would require │ │ │ └─ cudatoolkit 11.5|11.5. , which can be installed; │ │ ├─ cuda-version 11.6 would require │ │ │ └─ cudatoolkit 11.6|11.6. , which can be installed; │ │ └─ cuda-version 11.7 would require │ │ └─ cudatoolkit 11.7|11.7.* , which can be installed; │ └─ pytorch 2.5.1 cuda118303, which can be installed; ├─ cudatoolkit 10.2** is not installable because it conflicts with any installable versions previously reported; └─ pytorch is not installable because there are no viable options ├─ pytorch [0.4.0|0.4.1|...|2.5.1] conflicts with any installable versions previously reported; ├─ pytorch [1.10.0|1.10.1|...|1.9.1] would require │ └─ cudatoolkit >=11.1,<11.2 , which does not exist (perhaps a missing channel); ├─ pytorch 1.11.0 would require │ └─ cudatoolkit >=11.5,<11.6 , which does not exist (perhaps a missing channel); ├─ pytorch [1.12.0|1.12.1] would require │ └─ cudatoolkit >=11.6,<11.7 , which does not exist (perhaps a missing channel); ├─ pytorch [1.13.0|1.13.1] would require │ └─ pytorch-cuda >=11.6,<11.7 , which requires │ └─ cuda 11.6. , which does not exist (perhaps a missing channel); ├─ pytorch [1.13.0|1.13.1|2.0.0|2.0.1] would require │ └─ pytorch-cuda >=11.7,<11.8 but there are no viable options │ ├─ pytorch-cuda 11.7 would require │ │ └─ cuda-cudart >=11.7,<11.8 , which does not exist (perhaps a missing channel); │ └─ pytorch-cuda 11.7 would require │ └─ cuda 11.7. , which does not exist (perhaps a missing channel); ├─ pytorch [1.7.1|1.8.0] would require │ └─ python_abi 3.9. _cp39, which does not exist (perhaps a missing channel); ├─ pytorch [2.0.0|2.0.1|...|2.5.1] would require │ └─ pytorch-cuda >=11.8,<11.9 but there are no viable options │ ├─ pytorch-cuda 11.8 would require │ │ └─ cuda-cudart >=11.8,<12.0 , which does not exist (perhaps a missing channel); │ └─ pytorch-cuda 11.8 would require │ └─ cuda 11.8. , which does not exist (perhaps a missing channel); ├─ pytorch [2.1.0|2.1.1|...|2.5.1] would require │ └─ pytorch-cuda >=12.1,<12.2 , which requires │ └─ cuda-cudart >=12.1,<12.2 , which does not exist (perhaps a missing channel); ├─ pytorch [0.2.0|0.3.0] would require │ └─ cudatoolkit 7.5 , which does not exist (perhaps a missing channel); ├─ pytorch [0.2.0|0.3.0|0.3.1] would require │ └─ cudatoolkit 8.0 , which does not exist (perhaps a missing channel); ├─ pytorch [0.3.1|0.4.0] would require │ └─ cudatoolkit 8.0.* , which does not exist (perhaps a missing channel); ├─ pytorch 1.0.1 would require │ └─ cudatoolkit >=8.0,<8.1.0a0 , which does not exist (perhaps a missing channel); └─ pytorch [2.0.1|2.3.0] would require └─ cuda >=11.8 , which is missing on the system.

Ahh, I’ve made a mistake. For cpu version, try

conda install pytorch torchvision torchaudio cpuonly -c pytorch

Thanks, I installed abacus+deepks following the instructions you provided!