Re: [Ifeffit] Larch version up problem

7 Feb 2024

      Hello Matt

Thank you for your kindness.

I simplified the ff2chi problem in multiprocessing. I found that the ff2chi
problem in multiprocessing is caused by receiving the arguments in an
object (TestClass).

For instance
The below code works in multiprocessing
---------------------------------------------
class TestClass():
    def __init__(self, out_path_lists):
        self.path_lists=out_path_lists

    def test(self):
        path1 = feffpath(os.path.join(dPath,'FEFF1', fname1))
        path_lists=[path1]
        theory_sum=Group(label='theory_group_sum')
        ff2chi(path_lists, group=theory_sum)
        new_state=theory_sum.ch

However, If I change ff2chi(path_lists, group=theory_sum) to*
ff2chi(self.path_lists, group=theory_sum*), the ff2chi problem is caused.

Finally, I want to use multiprocessing. The xraylarch you sent does not
work in multiprocessing.
If you are possible, please send me larch 0.9.71 to reinstall

2024년 2월 7일 (수) 오전 7:37, Matt Newville 님이 작성:
...
Hi Eun-Suk,
Did you do
       python -m pip install
https://millenia.cars.aps.anl.gov/xraylarch/downloads/xraylarch-latest-py3-n...
?
That will install the development version (later than 0.9.74, not yet
0.9.75, sorry for that delay).  This works for me:
from larch.xafs import feffpath
  p1 = feffpath('feff0001.dat', e0=-1, sigma2=0.010)
  p1.s02 = 0.95
  dat = ff2chi([p1])
I don't know of any multiprocessing problem...
Cheers,
On Tue, Feb 6, 2024 at 11:05 PM 정은석  wrote:
...
Hello  Matt.
I think that you need some time to fix the 'multi-processing problem of
0. 9.74'. If you take a little long time to fix it, can you send me
xraylarch version 0.9.71? or could you inform me of the site of the old
version?
...
I will reinstall xraylarch with it.
Thank you for your effort and contribution
2024년 2월 5일 (월) 오후 5:09, 정은석 님이 작성:
...
Hello Matt
I changed my original code to simple code for you as below.
I found the problem of larch 0.9.74 in multi-processing.
You can check it with the below code. How can I fix this problem?

...
...
############# Multi-processing test ########################
import larch
from larch import Group
from larch.utils import *
from larch.xafs import *
from larch.io import *
import numpy as np
import os
import torch.multiprocessing as mp
dPath = r'C:\Users\Administrator\Desktop\RL\feffit_error_test'
os.chdir(dPath)
fname1 = 'feff0001.dat'  # N=1, Reff=1.9 N
path1 = feffpath(os.path.join(dPath,'FEFF1', fname1))
#################################
path_lists=[path1]
path_lists[0].s02=0.86
################################################
class Func():
    def __init__(self, path_lists):
        self.path_lists=path_lists
def step(self):
        self.path_lists[0].e0, self.path_lists[0].degen,
self.path_lists[0].deltar, self.path_lists[0].sigma2 = [2, 5, 0.001, 0.004]
    ## Theory path들의 합과 theory FFT
        theory_sum=Group(label='theory_group_sum')
        ff2chi(self.path_lists, group=theory_sum)
        new_state=theory_sum.chi
        return new_state
class TestClass():
    def __init__(self, path_lists):
        self.path_lists=path_lists
def test(self):
        func1 = Func(self.path_lists)
        s=func1.step()
############################################
T_func=TestClass(path_lists)
F_func=Func(path_lists)
if __name__ == '__main__':
    process_list=[]
    p_N =mp.Process(target=T_func.test,)
    p_N.start()
    process_list.append(p_N)
    for process in process_list:
        process.join()
##################################################################
#####Problem##########
  File
"C:\ProgramData\anaconda3\envs\xraylarch\Lib\multiprocessing\process.py",
line 314, in _bootstrap
    self.run()
  File
"C:\ProgramData\anaconda3\envs\xraylarch\Lib\multiprocessing\process.py",
line 108, in run
    self._target(*self._args, **self._kwargs)
  File
"c:\Users\Administrator\Desktop\RL\feffit_error_test\test-ff2chi-problem.py",
line 40, in test
    s=func1.step()
      ^^^^^^^^^^^^
  File
"c:\Users\Administrator\Desktop\RL\feffit_error_test\test-ff2chi-problem.py",
line 30, in step
    ff2chi(self.path_lists, group=theory_sum)
  File
"C:\ProgramData\anaconda3\envs\xraylarch\Lib\site-packages\larch\xafs\feffdat.py",
line 658, in ff2chi
    path.create_path_params(params=params)
  File
"C:\ProgramData\anaconda3\envs\xraylarch\Lib\site-packages\larch\xafs\feffdat.py",
line 392, in create_path_params
    parname = self.pathpar_name(pname)
              ^^^^^^^^^^^^^^^^^^^^^^^^
  File
"C:\ProgramData\anaconda3\envs\xraylarch\Lib\site-packages\larch\xafs\feffdat.py",
line 331, in pathpar_name
    return f'{parname}_{self.dataset}_{self.hashkey}'
                        ^^^^^^^^^^^^
AttributeError: 'FeffPathGroup' object has no attribute 'dataset'
2024년 2월 5일 (월) 오전 4:44, Matt Newville 님이
작성:
...
Hi Eun-Suk,
All the examples with ff2chi and running feffit work for me.   I sort
of don't understand how that could happen (a FeffPathGroup does -- or
should -- have a `dataset` attribute).
Can you provide an example that shows the problem?
--Matt
_______________________________________________
Ifeffit mailing list
Ifeffit@millenia.cars.aps.anl.gov
http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit
Unsubscribe: http://millenia.cars.aps.anl.gov/mailman/options/ifeffit
--
Best regards,
Ph. D. Eun-Suk Jeong
X-ray absorption fine structure(EXAFS+XANES)
Mobile:+82-10-4628-9896
--
Best regards,
Ph. D. Eun-Suk Jeong
X-ray absorption fine structure(EXAFS+XANES)
Mobile:+82-10-4628-9896
_______________________________________________
Ifeffit mailing list
Ifeffit@millenia.cars.aps.anl.gov
http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit
Unsubscribe: http://millenia.cars.aps.anl.gov/mailman/options/ifeffit
--
--Matt Newville <newville at cars.uchicago.edu> 630-327-7411
_______________________________________________
Ifeffit mailing list
Ifeffit@millenia.cars.aps.anl.gov
http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit
Unsubscribe: http://millenia.cars.aps.anl.gov/mailman/options/ifeffit
-- 
Best regards,

Ph. D. Eun-Suk Jeong
X-ray absorption fine structure(EXAFS+XANES)
Mobile:+82-10-4628-9896