Error with dataclass in queue submission (?)

[ligerzero-ai]

Some kind of weird error when I try to submit vasp to the queue

from pymatgen.io.vasp.inputs import Incar
from ase.build import bulk
from pymatgen.io.ase import AseAtomsAdaptor
#from vasp_nodes import get_default_POTCAR_paths, write_POTCAR, VaspInput, vasp_job, create_WorkingDirectory, stack_element_string
import numpy as np
from pyiron_workflow import for_node, Workflow

%load_ext autoreload
%autoreload 2

from pyiron_nodes.atomistic.engine.vasp import get_default_POTCAR_paths, write_POTCAR, VaspInput, vasp_job, create_WorkingDirectory, stack_element_string

incar = Incar.from_dict({
    "ENCUT": 520,
    "EDIFF": 1e-4,
    "ISMEAR": 1,
    "SIGMA": 0.1,
})

bulk_Fe = bulk("Fe", cubic=True, a=2.83)
bulk_Fe = AseAtomsAdaptor().get_structure(bulk_Fe)
vi = VaspInput(bulk_Fe, incar, potcar_paths=["/cmmc/u/hmai/vasp_potentials_54/Fe_sv/POTCAR"])

import pyiron_workflow as pwf

def submit_to_slurm(
    node,
    /,
    job_name=None,
    output_file=None,
    error_file=None,
    time_limit="00:05:00",
    account="hmai",
    partition="s.cmmg",
    nodes=1,
    ntasks=40,
    cpus_per_task=40,
    memory="1GB",
):
    """
    An example of a helper function for running nodes on slurm.

    - Saves the node
    - Writes a slurm batch script that 
        - Loads the node
        - Runs it
        - Saves it again
    - Runs the batch script
    """
    if node.graph_root is not node:
        raise ValueError(
            f"Can only submit parent-most nodes, but {node.full_label} "
            f"has root {node.graph_root.full_label}"
        )
        
    node.save(backend="pickle")
    p = node.as_path()
    
    if job_name is None:
        job_name = node.full_label 
        job_name = job_name.replace(node.semantic_delimiter, "_")
        job_name = "pwf" + job_name
        
    script_content = f"""#!/bin/bash
#SBATCH --partition=s.cmmg
#SBATCH --ntasks=1  # Adjust CPU count as needed
#SBATCH --cpus-per-task=1
#SBATCH --time=10:00:00  # Adjust wall time as needed
#SBATCH --output=time.out
#SBATCH --error=error.out
#SBATCH --job-name=resubmitter.sh  # Adjust job name as needed
#SBATCH --get-user-env=L
#SBATCH --mem-per-cpu=2000MB
#SBATCH --hint=nomultithread
##SBATCH --reservation=benchmarking

# Execute Python script inline
python - <<EOF
from pyiron_workflow import PickleStorage
node = PickleStorage().load(filename="{node.as_path().joinpath('picklestorage').resolve()}")  # Load
node.run()  # Run
node.save(backend="pickle")  # Save again
EOF
"""
    submission_script = p.joinpath("node_submission.sh")
    submission_script.write_text(script_content)
    import subprocess
    submission = subprocess.run(["sbatch", submission_script.resolve()])
    return submission

#v_job = vasp_job(workdir="/cmmc/u/hmai/github_dev_pyiron/pyiron_vasp_nodes/test3", vasp_input = vi)
wf = pwf.Workflow("vasp_queue")
wf.vasp_job_queue = vasp_job(workdir="/cmmc/u/hmai/github_dev_pyiron/pyiron_vasp_nodes/test3",
                             vasp_input = vi)
# Send it off
submit_to_slurm(wf)

# Try reloading it (need to wait for `submit_to_slurm`)
while not wf.has_saved_content():
    sleep(0.2)
reloaded = pwf.Workflow(wf.label)

# Reload it until it's done
while reloaded.vasp_job_queue.outputs is pwf.NOT_DATA:
    reloaded.load()
    print(reloaded.vasp_job_queue.outputs.to_value_dict()["vasp_output"]["generic"]["energy_pot"])
    sleep(1)
reloaded.outputs.to_value_dict()

In the error.out

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/cmmc/ptmp/pyironhb/mambaforge/envs/pyiron_mpie_cmti_2024-11-18/lib/python3.11/site-packages/pyiron_workflow/__init__.py", line 37, in <module>
    from pyiron_workflow.workflow import Workflow
  File "/cmmc/ptmp/pyironhb/mambaforge/envs/pyiron_mpie_cmti_2024-11-18/lib/python3.11/site-packages/pyiron_workflow/workflow.py", line 13, in <module>
    from pyiron_workflow.nodes.composite import Composite
  File "/cmmc/ptmp/pyironhb/mambaforge/envs/pyiron_mpie_cmti_2024-11-18/lib/python3.11/site-packages/pyiron_workflow/nodes/composite.py", line 15, in <module>
    from pyiron_workflow.create import HasCreator
  File "/cmmc/ptmp/pyironhb/mambaforge/envs/pyiron_mpie_cmti_2024-11-18/lib/python3.11/site-packages/pyiron_workflow/create.py", line 16, in <module>
    from pyiron_workflow.nodes.function import function_node, as_function_node
  File "/cmmc/ptmp/pyironhb/mambaforge/envs/pyiron_mpie_cmti_2024-11-18/lib/python3.11/site-packages/pyiron_workflow/nodes/function.py", line 10, in <module>
    from pyiron_workflow.mixin.preview import ScrapesIO
  File "/cmmc/ptmp/pyironhb/mambaforge/envs/pyiron_mpie_cmti_2024-11-18/lib/python3.11/site-packages/pyiron_workflow/mixin/preview.py", line 28, in <module>
    from pyiron_workflow.channels import NOT_DATA
  File "/cmmc/ptmp/pyironhb/mambaforge/envs/pyiron_mpie_cmti_2024-11-18/lib/python3.11/site-packages/pyiron_workflow/channels.py", line 19, in <module>
    from pyiron_workflow.type_hinting import (
  File "/cmmc/ptmp/pyironhb/mambaforge/envs/pyiron_mpie_cmti_2024-11-18/lib/python3.11/site-packages/pyiron_workflow/type_hinting.py", line 10, in <module>
    from pint import Quantity
  File "/cmmc/ptmp/pyironhb/mambaforge/envs/pyiron_mpie_cmti_2024-11-18/lib/python3.11/site-packages/pint/__init__.py", line 18, in <module>
    from .delegates.formatter._format_helpers import formatter
  File "/cmmc/ptmp/pyironhb/mambaforge/envs/pyiron_mpie_cmti_2024-11-18/lib/python3.11/site-packages/pint/delegates/__init__.py", line 12, in <module>
    from . import txt_defparser
  File "/cmmc/ptmp/pyironhb/mambaforge/envs/pyiron_mpie_cmti_2024-11-18/lib/python3.11/site-packages/pint/delegates/txt_defparser/__init__.py", line 12, in <module>
    from .defparser import DefParser
  File "/cmmc/ptmp/pyironhb/mambaforge/envs/pyiron_mpie_cmti_2024-11-18/lib/python3.11/site-packages/pint/delegates/txt_defparser/defparser.py", line 10, in <module>
    from . import block, common, context, defaults, group, plain, system
  File "/cmmc/ptmp/pyironhb/mambaforge/envs/pyiron_mpie_cmti_2024-11-18/lib/python3.11/site-packages/pint/delegates/txt_defparser/common.py", line 23, in <module>
    @dataclass(frozen=True)
     ^^^^^^^^^^^^^^^^^^^^^^
  File "/cmmc/ptmp/pyironhb/mambaforge/envs/pyiron_mpie_cmti_2024-11-18/lib/python3.11/dataclasses.py", line 1220, in wrap
    return _process_class(cls, init, repr, eq, order, unsafe_hash,
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/cmmc/ptmp/pyironhb/mambaforge/envs/pyiron_mpie_cmti_2024-11-18/lib/python3.11/dataclasses.py", line 993, in _process_class
    raise TypeError('cannot inherit frozen dataclass from a '
TypeError: cannot inherit frozen dataclass from a non-frozen one

[jan-janssen]

That seems to be a pint related issue hgrecco/pint#1969 (comment) - @niklassiemer is already working on it pyiron/docker-stacks#417

[liamhuber]

I hope that's it, because otherwise I am not seeing how to fix it.

On the off chance that this isn't purely an upstream/pint issue and we need to dig deeper, try simplifying your example like this:

import pyiron_workflow as pwf
from dataclasses import dataclass

@dataclass
class SomeData:
    foo: str = "foo"
    bar: int = 42

@pwf.as_function_node
def SomeNode(data):
    return data

n = SomeNode(SomeData())

# Then the rest of the submission and load looping on this simpler node

[niklassiemer]

If it is the pint stuff, it will break by just import pint and, thus, also on import pyiron_workflow. The resulting error is for sure very similar!

[liamhuber]

Right, and the import line itself is indeed at the root of the stack trace, I should have noticed that. Looking super probable this is a non-workflow issue, and hopefully the new release with the latest pint will resolve the downgrade issue you're having on the docker stack

[ligerzero-ai]

I fixed (?) it by running in a completely separate environment with my own pyiron_workflow.

The good news is that the first vasp calculation using liam's method of node serialisation was now successful!

Looking forward to running production