Mercurial > repos > bgruening > keras_batch_models
comparison simple_model_fit.py @ 3:0a7f2b9e1fcb draft
"planemo upload for repository https://github.com/bgruening/galaxytools/tree/master/tools/sklearn commit eb703290e2589561ea215c84aa9f71bcfe1712c6"
author | bgruening |
---|---|
date | Fri, 01 Nov 2019 17:20:47 -0400 |
parents | |
children | 9afec58b2bf9 |
comparison
equal
deleted
inserted
replaced
2:7f6079deac6f | 3:0a7f2b9e1fcb |
---|---|
1 import argparse | |
2 import json | |
3 import pandas as pd | |
4 import pickle | |
5 | |
6 from galaxy_ml.utils import load_model, read_columns | |
7 from sklearn.pipeline import Pipeline | |
8 | |
9 | |
10 def _get_X_y(params, infile1, infile2): | |
11 """ read from inputs and output X and y | |
12 | |
13 Parameters | |
14 ---------- | |
15 params : dict | |
16 Tool inputs parameter | |
17 infile1 : str | |
18 File path to dataset containing features | |
19 infile2 : str | |
20 File path to dataset containing target values | |
21 | |
22 """ | |
23 # store read dataframe object | |
24 loaded_df = {} | |
25 | |
26 input_type = params['input_options']['selected_input'] | |
27 # tabular input | |
28 if input_type == 'tabular': | |
29 header = 'infer' if params['input_options']['header1'] else None | |
30 column_option = (params['input_options']['column_selector_options_1'] | |
31 ['selected_column_selector_option']) | |
32 if column_option in ['by_index_number', 'all_but_by_index_number', | |
33 'by_header_name', 'all_but_by_header_name']: | |
34 c = params['input_options']['column_selector_options_1']['col1'] | |
35 else: | |
36 c = None | |
37 | |
38 df_key = infile1 + repr(header) | |
39 df = pd.read_csv(infile1, sep='\t', header=header, | |
40 parse_dates=True) | |
41 loaded_df[df_key] = df | |
42 | |
43 X = read_columns(df, c=c, c_option=column_option).astype(float) | |
44 # sparse input | |
45 elif input_type == 'sparse': | |
46 X = mmread(open(infile1, 'r')) | |
47 | |
48 # Get target y | |
49 header = 'infer' if params['input_options']['header2'] else None | |
50 column_option = (params['input_options']['column_selector_options_2'] | |
51 ['selected_column_selector_option2']) | |
52 if column_option in ['by_index_number', 'all_but_by_index_number', | |
53 'by_header_name', 'all_but_by_header_name']: | |
54 c = params['input_options']['column_selector_options_2']['col2'] | |
55 else: | |
56 c = None | |
57 | |
58 df_key = infile2 + repr(header) | |
59 if df_key in loaded_df: | |
60 infile2 = loaded_df[df_key] | |
61 else: | |
62 infile2 = pd.read_csv(infile2, sep='\t', | |
63 header=header, parse_dates=True) | |
64 loaded_df[df_key] = infile2 | |
65 | |
66 y = read_columns( | |
67 infile2, | |
68 c=c, | |
69 c_option=column_option, | |
70 sep='\t', | |
71 header=header, | |
72 parse_dates=True) | |
73 if len(y.shape) == 2 and y.shape[1] == 1: | |
74 y = y.ravel() | |
75 | |
76 return X, y | |
77 | |
78 | |
79 def main(inputs, infile_estimator, infile1, infile2, out_object, | |
80 out_weights=None): | |
81 """ main | |
82 | |
83 Parameters | |
84 ---------- | |
85 inputs : str | |
86 File path to galaxy tool parameter | |
87 | |
88 infile_estimator : str | |
89 File paths of input estimator | |
90 | |
91 infile1 : str | |
92 File path to dataset containing features | |
93 | |
94 infile2 : str | |
95 File path to dataset containing target labels | |
96 | |
97 out_object : str | |
98 File path for output of fitted model or skeleton | |
99 | |
100 out_weights : str | |
101 File path for output of weights | |
102 | |
103 """ | |
104 with open(inputs, 'r') as param_handler: | |
105 params = json.load(param_handler) | |
106 | |
107 # load model | |
108 with open(infile_estimator, 'rb') as est_handler: | |
109 estimator = load_model(est_handler) | |
110 | |
111 X_train, y_train = _get_X_y(params, infile1, infile2) | |
112 | |
113 estimator.fit(X_train, y_train) | |
114 | |
115 main_est = estimator | |
116 if isinstance(main_est, Pipeline): | |
117 main_est = main_est.steps[-1][-1] | |
118 if hasattr(main_est, 'model_') \ | |
119 and hasattr(main_est, 'save_weights'): | |
120 if out_weights: | |
121 main_est.save_weights(out_weights) | |
122 del main_est.model_ | |
123 del main_est.fit_params | |
124 del main_est.model_class_ | |
125 del main_est.validation_data | |
126 if getattr(main_est, 'data_generator_', None): | |
127 del main_est.data_generator_ | |
128 | |
129 with open(out_object, 'wb') as output_handler: | |
130 pickle.dump(estimator, output_handler, | |
131 pickle.HIGHEST_PROTOCOL) | |
132 | |
133 | |
134 if __name__ == '__main__': | |
135 aparser = argparse.ArgumentParser() | |
136 aparser.add_argument("-i", "--inputs", dest="inputs", required=True) | |
137 aparser.add_argument("-X", "--infile_estimator", dest="infile_estimator") | |
138 aparser.add_argument("-y", "--infile1", dest="infile1") | |
139 aparser.add_argument("-g", "--infile2", dest="infile2") | |
140 aparser.add_argument("-o", "--out_object", dest="out_object") | |
141 aparser.add_argument("-t", "--out_weights", dest="out_weights") | |
142 args = aparser.parse_args() | |
143 | |
144 main(args.inputs, args.infile_estimator, args.infile1, | |
145 args.infile2, args.out_object, args.out_weights) |