Mercurial > repos > bgruening > stacking_ensemble_models

comparison stacking_ensembles.py @ 3:0a1812986bc3 draft

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
planemo upload for repository https://github.com/bgruening/galaxytools/tree/master/tools/sklearn commit 9981e25b00de29ed881b2229a173a8c812ded9bb
author bgruening
date Wed, 09 Aug 2023 11:10:37 +0000
parents 38c4f8a98038
children
comparison
equal deleted inserted replaced
2:38c4f8a98038 3:0a1812986bc3
1 import argparse 1 import argparse
2 import ast 2 import ast
3 import json 3 import json
4 import mlxtend.regressor
5 import mlxtend.classifier
6 import pandas as pd
7 import pickle
8 import sklearn
9 import sys 4 import sys
10 import warnings 5 import warnings
11 from sklearn import ensemble 6 from distutils.version import LooseVersion as Version
12 7
13 from galaxy_ml.utils import (load_model, get_cv, get_estimator, 8 import mlxtend.classifier
14 get_search_params) 9 import mlxtend.regressor
10 from galaxy_ml import __version__ as galaxy_ml_version
11 from galaxy_ml.model_persist import dump_model_to_h5, load_model_from_h5
12 from galaxy_ml.utils import get_cv, get_estimator
13
14 warnings.filterwarnings("ignore")
15
16 N_JOBS = int(__import__("os").environ.get("GALAXY_SLOTS", 1))
15 17
16 18
17 warnings.filterwarnings('ignore') 19 def main(inputs_path, output_obj, base_paths=None, meta_path=None):
18
19 N_JOBS = int(__import__('os').environ.get('GALAXY_SLOTS', 1))
20
21
22 def main(inputs_path, output_obj, base_paths=None, meta_path=None,
23 outfile_params=None):
24 """ 20 """
25 Parameter 21 Parameter
26 --------- 22 ---------
27 inputs_path : str 23 inputs_path : str
28 File path for Galaxy parameters 24 File path for Galaxy parameters
33 base_paths : str 29 base_paths : str
34 File path or paths concatenated by comma. 30 File path or paths concatenated by comma.
35 31
36 meta_path : str 32 meta_path : str
37 File path 33 File path
38
39 outfile_params : str
40 File path for params output
41 """ 34 """
42 with open(inputs_path, 'r') as param_handler: 35 with open(inputs_path, "r") as param_handler:
43 params = json.load(param_handler) 36 params = json.load(param_handler)
44 37
45 estimator_type = params['algo_selection']['estimator_type'] 38 estimator_type = params["algo_selection"]["estimator_type"]
46 # get base estimators 39 # get base estimators
47 base_estimators = [] 40 base_estimators = []
48 for idx, base_file in enumerate(base_paths.split(',')): 41 for idx, base_file in enumerate(base_paths.split(",")):
49 if base_file and base_file != 'None': 42 if base_file and base_file != "None":
50 with open(base_file, 'rb') as handler: 43 model = load_model_from_h5(base_file)
51 model = load_model(handler)
52 else: 44 else:
53 estimator_json = (params['base_est_builder'][idx] 45 estimator_json = params["base_est_builder"][idx]["estimator_selector"]
54 ['estimator_selector'])
55 model = get_estimator(estimator_json) 46 model = get_estimator(estimator_json)
56 47
57 if estimator_type.startswith('sklearn'): 48 if estimator_type.startswith("sklearn"):
58 named = model.__class__.__name__.lower() 49 named = model.__class__.__name__.lower()
59 named = 'base_%d_%s' % (idx, named) 50 named = "base_%d_%s" % (idx, named)
60 base_estimators.append((named, model)) 51 base_estimators.append((named, model))
61 else: 52 else:
62 base_estimators.append(model) 53 base_estimators.append(model)
63 54
64 # get meta estimator, if applicable 55 # get meta estimator, if applicable
65 if estimator_type.startswith('mlxtend'): 56 if estimator_type.startswith("mlxtend"):
66 if meta_path: 57 if meta_path:
67 with open(meta_path, 'rb') as f: 58 meta_estimator = load_model_from_h5(meta_path)
68 meta_estimator = load_model(f)
69 else: 59 else:
70 estimator_json = (params['algo_selection'] 60 estimator_json = params["algo_selection"]["meta_estimator"][
71 ['meta_estimator']['estimator_selector']) 61 "estimator_selector"
62 ]
72 meta_estimator = get_estimator(estimator_json) 63 meta_estimator = get_estimator(estimator_json)
73 64
74 options = params['algo_selection']['options'] 65 options = params["algo_selection"]["options"]
75 66
76 cv_selector = options.pop('cv_selector', None) 67 cv_selector = options.pop("cv_selector", None)
77 if cv_selector: 68 if cv_selector:
69 if Version(galaxy_ml_version) < Version("0.8.3"):
70 cv_selector.pop("n_stratification_bins", None)
78 splitter, groups = get_cv(cv_selector) 71 splitter, groups = get_cv(cv_selector)
79 options['cv'] = splitter 72 options["cv"] = splitter
80 # set n_jobs 73 # set n_jobs
81 options['n_jobs'] = N_JOBS 74 options["n_jobs"] = N_JOBS
82 75
83 weights = options.pop('weights', None) 76 weights = options.pop("weights", None)
84 if weights: 77 if weights:
85 weights = ast.literal_eval(weights) 78 weights = ast.literal_eval(weights)
86 if weights: 79 if weights:
87 options['weights'] = weights 80 options["weights"] = weights
88 81
89 mod_and_name = estimator_type.split('_') 82 mod_and_name = estimator_type.split("_")
90 mod = sys.modules[mod_and_name[0]] 83 mod = sys.modules[mod_and_name[0]]
91 klass = getattr(mod, mod_and_name[1]) 84 klass = getattr(mod, mod_and_name[1])
92 85
93 if estimator_type.startswith('sklearn'): 86 if estimator_type.startswith("sklearn"):
94 options['n_jobs'] = N_JOBS 87 options["n_jobs"] = N_JOBS
95 ensemble_estimator = klass(base_estimators, **options) 88 ensemble_estimator = klass(base_estimators, **options)
96 89
97 elif mod == mlxtend.classifier: 90 elif mod == mlxtend.classifier:
98 ensemble_estimator = klass( 91 ensemble_estimator = klass(
99 classifiers=base_estimators, 92 classifiers=base_estimators, meta_classifier=meta_estimator, **options
100 meta_classifier=meta_estimator, 93 )
101 **options)
102 94
103 else: 95 else:
104 ensemble_estimator = klass( 96 ensemble_estimator = klass(
105 regressors=base_estimators, 97 regressors=base_estimators, meta_regressor=meta_estimator, **options
106 meta_regressor=meta_estimator, 98 )
107 **options)
108 99
109 print(ensemble_estimator) 100 print(ensemble_estimator)
110 for base_est in base_estimators: 101 for base_est in base_estimators:
111 print(base_est) 102 print(base_est)
112 103
113 with open(output_obj, 'wb') as out_handler: 104 dump_model_to_h5(ensemble_estimator, output_obj)
114 pickle.dump(ensemble_estimator, out_handler, pickle.HIGHEST_PROTOCOL)
115
116 if params['get_params'] and outfile_params:
117 results = get_search_params(ensemble_estimator)
118 df = pd.DataFrame(results, columns=['', 'Parameter', 'Value'])
119 df.to_csv(outfile_params, sep='\t', index=False)
120 105
121 106
122 if __name__ == '__main__': 107 if __name__ == "__main__":
123 aparser = argparse.ArgumentParser() 108 aparser = argparse.ArgumentParser()
124 aparser.add_argument("-b", "--bases", dest="bases") 109 aparser.add_argument("-b", "--bases", dest="bases")
125 aparser.add_argument("-m", "--meta", dest="meta") 110 aparser.add_argument("-m", "--meta", dest="meta")
126 aparser.add_argument("-i", "--inputs", dest="inputs") 111 aparser.add_argument("-i", "--inputs", dest="inputs")
127 aparser.add_argument("-o", "--outfile", dest="outfile") 112 aparser.add_argument("-o", "--outfile", dest="outfile")
128 aparser.add_argument("-p", "--outfile_params", dest="outfile_params")
129 args = aparser.parse_args() 113 args = aparser.parse_args()
130 114
131 main(args.inputs, args.outfile, base_paths=args.bases, 115 main(args.inputs, args.outfile, base_paths=args.bases, meta_path=args.meta)
132 meta_path=args.meta, outfile_params=args.outfile_params)