Can't restart active learning trajectories

[FromColleWithLove]

Describe the bug
Cannot restart active learning trajectories

To Reproduce
Complete an active learning trajectory, then restart it by running, in the folder containing all files:

flare-otf restart.yaml

Where restart.yaml is:

otf:
        mode: restart
        number_of_steps : 250000
        checkpoint: Au_otf_checkpt.json

The active learning goes on until it needs to call DFT, at which point, instead of performing ab initio calculations, it raise this error:

Traceback (most recent call last):
  File "[...]/miniconda3/envs/envflare/bin/flare-otf", line 8, in <module>
    sys.exit(main())
             ^^^^^^
  File "[...]/miniconda3/envs/envflare/lib/python3.12/site-packages/flare/scripts/otf_train.py", line 380, in main
    restart_otf(config)
  File "[...]/miniconda3/envs/envflare/lib/python3.12/site-packages/flare/scripts/otf_train.py", line 369, in restart_otf
    otf.run()
  File "[...]/miniconda3/envs/envflare/lib/python3.12/site-packages/flare/learners/otf.py", line 442, in run
    self.md_step()  # update positions by Verlet
    ^^^^^^^^^^^^^^
  File "[...]/miniconda3/envs/envflare/lib/python3.12/site-packages/flare/learners/otf.py", line 547, in md_step
    self.md.step(tol, self.number_of_steps)
  File "[...]/miniconda3/envs/envflare/lib/python3.12/site-packages/flare/md/lammps.py", line 297, in step
    self.backup(trj)
  File "[...]/miniconda3/envs/envflare/lib/python3.12/site-packages/flare/md/lammps.py", line 319, in backup
    thermostat.shape[0]
AssertionError

Package version:
mir-flare 1.4.2

Additional context
The error occurred using 48 cores on a node as well as using 192 cores on 4 nodes.