version of joblib?
Closed this issue · 3 comments
I have installed joblib=1.0.1 in pyhon3.6, but when I run run.py in python3.6, 'ImportError' is still there.
Validating method 1/1: "sp_sg"
['phototourism', 'pragueparks', 'googleurban']
Running: phototourism, stereo track
Running: phototourism, multiview track
Running: pragueparks, stereo track
Running: pragueparks, multiview track
Running: googleurban, stereo track
Running: googleurban, multiview track
Validating key "config_phototourism_stereo"
Validating key "config_phototourism_multiview"
Validating key "config_pragueparks_stereo"
Validating key "config_pragueparks_multiview"
Validating key "config_googleurban_stereo"
Validating key "config_googleurban_multiview"
Working on sp_sg: phototourism/reichstag
-- File feature already exists
-- Computing match
WARNING: ./jobs/9efa086bd2840045bf66679f8d8ad4a21a0d23201b45937cb4c741cf8aa6e00d already exists!
Traceback (most recent call last):
File "compute_match.py", line 19, in
from joblib import Parallel, delayed
ImportError: No module named joblib
Traceback (most recent call last):
File "run.py", line 280, in
main(cfg)
File "run.py", line 219, in main
job_dict)
File "run.py", line 47, in create_eval_jobs
job = create_and_queue_jobs(cmd_list, cfg, dep_str)
File "/cephfs/person/dihehuang/imc-2021-submit/image-matching-benchmark/utils/queue_helper.py", line 310, in create_and_queue_jobs
cpu=cpu)
File "/cephfs/person/dihehuang/imc-2021-submit/image-matching-benchmark/utils/queue_helper.py", line 161, in queue_job
raise RuntimeError('Subprocess error!')
RuntimeError: Subprocess error!
That's probably an error in the job itself, not joblib. Check logs/9efa086bd2840045bf66679f8d8ad4a21a0d23201b45937cb4c741cf8aa6e00d.log or, easier, just re-run the pipeline with the --run_mode=interactive flag to figure out why it's crashing.
That's probably an error in the job itself, not joblib. Check logs/9efa086bd2840045bf66679f8d8ad4a21a0d23201b45937cb4c741cf8aa6e00d.log or, easier, just re-run the pipeline with the
--run_mode=interactiveflag to figure out why it's crashing.
I have set '--run_mode=interactive', but there is no log file in ./logs
when I set sys.path.append('/usr/local/lib/python3.6/site-packages')
it occurs:
Traceback (most recent call last):
File "compute_match.py", line 20, in
from joblib import Parallel, delayed
File "/usr/local/lib/python3.6/site-packages/joblib/init.py", line 113, in
from .memory import Memory, MemorizedResult, register_store_backend
File "/usr/local/lib/python3.6/site-packages/joblib/memory.py", line 274
raise new_exc from exc
^
SyntaxError: invalid syntax
Traceback (most recent call last):
File "run.py", line 281, in
main(cfg)
File "run.py", line 220, in main
job_dict)
File "run.py", line 48, in create_eval_jobs
job = create_and_queue_jobs(cmd_list, cfg, dep_str)
File "/cephfs/person/dihehuang/imc-2021-submit/image-matching-benchmark/utils/queue_helper.py", line 310, in create_and_queue_jobs
cpu=cpu)
File "/cephfs/person/dihehuang/imc-2021-submit/image-matching-benchmark/utils/queue_helper.py", line 161, in queue_job
raise RuntimeError('Subprocess error!')
RuntimeError: Subprocess error!
Oh, I see your error message was indeed related to joblib, my bad. I presume this also happens when you import the library from the REPL? It's not a benchmark issue, and it's impossible for us to diagnose it. I would try installing joblib in a new conda environment and if that works, install the rest of the dependencies while ignoring library versions.
(Also: the logs only show up if you use --run_mode=batch, if you set it to interactive, you see error messages in the terminal.)