Repository navigation
Running regression flow requires clases? #312
Description
See example error below.
%matplotlib inline
import openml as oml
import pandas as pd
import numpy as np
import seaborn as sns
from matplotlib import pyplot
from sklearn import neighbors, model_selection, tree, ensemble
oml.config.apikey = "5a17d410c0a181e5ec2595ee9da79849"
task = oml.tasks.get_task(146685)
clf = ensemble.RandomForestRegressor()
flow = oml.flows.sklearn_to_flow(clf)
run = oml.runs.run_flow_on_task(task, flow)
#myrun = run.publish()
#print("Uploaded to http://www.openml.org/r/" + str(myrun.run_id))
---------------------------------------------------------------------------
AttributeError Traceback (most recent call last)
<ipython-input-5-e4b1318833fe> in <module>()
15 flow = oml.flows.sklearn_to_flow(clf)
16
---> 17 run = oml.runs.run_flow_on_task(task, flow)
18 #myrun = run.publish()
19 #print("Uploaded to http://www.openml.org/r/" + str(myrun.run_id))
/usr/local/lib/python3.6/site-packages/openml/runs/functions.py in run_flow_on_task(task, flow, avoid_duplicate_runs, flow_tags, seed)
99
100 # execute the run
--> 101 res = _run_task_get_arffcontent(flow.model, task, class_labels)
102
103 if flow.flow_id is None:
/usr/local/lib/python3.6/site-packages/openml/runs/functions.py in _run_task_get_arffcontent(model, task, class_labels)
424 model_classes = used_estimator.best_estimator_.classes_
425 else:
--> 426 model_classes = used_estimator.classes_
427
428 if can_measure_runtime:
AttributeError: 'RandomForestRegressor' object has no attribute 'classes_'
Activity
mfeurer commentedon Sep 27, 2017
We didn't implement regression yet. Therefore, the actual issue is that the error message is a very bad one for 'sorry, this isn't implemented yet'.
joaquinvanschoren commentedon Sep 27, 2017
mfeurer commentedon Sep 29, 2017
I don't know, partly because of openml/OpenML#413. Probably one has to generalize this file for different task types.
joaquinvanschoren commentedon Mar 3, 2018
This really feels like a bug for users. Can we make this a priority for the next workshop? I'll try to look into it earlier.
I don't see how this is related to #openml/OpenML#413?
mfeurer commentedon Mar 5, 2018
Could you please comment on priorities in #410 ? It is related to openml/OpenML#413 because I wouldn't know how to upload the trace of a regression task.
joaquinvanschoren commentedon Mar 28, 2018
Yes, I think this is a priority.
See this recent zinger on reddit:
https://www.reddit.com/r/MachineLearning/comments/83ju64/d_other_sites_besides_kaggle_where_you_can/
mfeurer commentedon Feb 25, 2019
Fixed via #560