Error when running mace_active_learning_md in committee mode

[zhangyu0104]

Hi,

I am running mace_active_learning_md with four pre-trained models in committee mode, but the program reports an error while writing an traj.xyz file.

I am not sure what causes this problem, so I have attached my running command, initial structure, and full error message below for reference.

Environment:
python: 3.11.13
mace-torch: 0.3.14
ase: 3.25.0

Running Command:

mace_active_learning_md \
  --config="./conf/init.xyz" \
  --error_threshold="0.2" \
  --temperature_K="5000" \
  --friction="0.1" \
  --timestep="1" \
  --nsteps="1000" \
  --nprint="10" \
  --nsave="10" \
  --ncheckerror="10" \
  --model="./model/mace00_run-*_stagetwo.model" \
  --output="traj.xyz" \
  --device="cuda" \
  --default_dtype="float64"

Initial structure:

166
Lattice="8.645 0.0 0.0 0.0 18.018 0.0 0.0 0.0 8.645" Properties=species:S:1:pos:R:3 pbc="T T T"
Mg       2.32365607       1.15424289       4.58179031
Mg       8.25998208       3.59125969       5.52532406
...
O        3.58322064       5.57755445       1.02576205

Error Message:

Running committee mace with 4 models
Using head default out of ['default']
2.3s: Energy per atom: Epot = -6.190eV  Ekin = 0.638eV (T=4934K)  Etot = -5.552eV t=0.0fs Eerr = 4.745eV Ferr = 2.200eV/A
Traceback (most recent call last):
  File "/home/ess/zhangyu23/anaconda3/envs/mace/bin/mace_active_learning_md", line 7, in <module>
    sys.exit(main())
             ^^^^^^
  File "/home/ess/zhangyu23/anaconda3/envs/mace/lib/python3.11/site-packages/mace/cli/active_learning_md.py", line 145, in main
    run(args)
  File "/home/ess/zhangyu23/anaconda3/envs/mace/lib/python3.11/site-packages/mace/cli/active_learning_md.py", line 189, in run
    dyn.run(NSTEPS)
  File "/home/ess/zhangyu23/anaconda3/envs/mace/lib/python3.11/site-packages/ase/md/md.py", line 171, in run
    return Dynamics.run(self, steps=steps)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/ess/zhangyu23/anaconda3/envs/mace/lib/python3.11/site-packages/ase/optimize/optimize.py", line 286, in run
    for converged in Dynamics.irun(self, steps=steps):
  File "/home/ess/zhangyu23/anaconda3/envs/mace/lib/python3.11/site-packages/ase/optimize/optimize.py", line 244, in irun
    self.call_observers()
  File "/home/ess/zhangyu23/anaconda3/envs/mace/lib/python3.11/site-packages/ase/optimize/optimize.py", line 207, in call_observers
    function(*args, **kwargs)
  File "/home/ess/zhangyu23/anaconda3/envs/mace/lib/python3.11/site-packages/mace/cli/active_learning_md.py", line 123, in save_config
    ase.io.write(fname, atomsi, append=True)
  File "/home/ess/zhangyu23/anaconda3/envs/mace/lib/python3.11/site-packages/ase/io/formats.py", line 707, in write
    return _write(filename, fd, format, io, images,
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/ess/zhangyu23/anaconda3/envs/mace/lib/python3.11/site-packages/ase/parallel.py", line 274, in new_func
    return func(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^
  File "/home/ess/zhangyu23/anaconda3/envs/mace/lib/python3.11/site-packages/ase/io/formats.py", line 743, in _write
    return io.write(fd, images, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/ess/zhangyu23/anaconda3/envs/mace/lib/python3.11/site-packages/ase/io/formats.py", line 204, in _write_wrapper
    return function(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/ess/zhangyu23/anaconda3/envs/mace/lib/python3.11/site-packages/ase/utils/__init__.py", line 599, in iofunc
    obj = func(fd, *args, **kwargs)
          ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/ess/zhangyu23/anaconda3/envs/mace/lib/python3.11/site-packages/ase/io/extxyz.py", line 942, in write_xyz
    data[column] = np.squeeze(value)
    ~~~~^^^^^^^^
ValueError: could not broadcast input array from shape (4,) into shape (166,)

[ilyes319]

Hey @zhangyu0104 It is been a long time since i have upgraded this script so it must need to changes to work now. Will add to the todo.

[Wang-Zhiwei24]

Hello! @ilyes319 I'm also encountering a similar error. I was wondering if there has been any progress or update on fixing this bug?
For reference, my environment details are:
ase: 3.27.0
mace-torch: 0.3.12
Python: 3.11.14