Mercurial > repos > bgruening > sklearn_train_test_split

comparison keras_train_and_eval.py @ 6:81ab4951f2a3 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 ca87db9c038a6fcf96aa39da50f384865fd932ff"
author bgruening
date Tue, 20 Apr 2021 17:09:29 +0000
parents c0ed68e280a7
children 82f89e379413
comparison
equal deleted inserted replaced
5:c0ed68e280a7 6:81ab4951f2a3
9 import numpy as np 9 import numpy as np
10 import pandas as pd 10 import pandas as pd
11 from galaxy_ml.externals.selene_sdk.utils import compute_score 11 from galaxy_ml.externals.selene_sdk.utils import compute_score
12 from galaxy_ml.keras_galaxy_models import _predict_generator 12 from galaxy_ml.keras_galaxy_models import _predict_generator
13 from galaxy_ml.model_validations import train_test_split 13 from galaxy_ml.model_validations import train_test_split
14 from galaxy_ml.utils import ( 14 from galaxy_ml.utils import (clean_params, get_main_estimator,
15 clean_params, 15 get_module, get_scoring, load_model, read_columns,
16 get_main_estimator, 16 SafeEval, try_get_attr)
17 get_module,
18 get_scoring,
19 load_model,
20 read_columns,
21 SafeEval,
22 try_get_attr,
23 )
24 from scipy.io import mmread 17 from scipy.io import mmread
25 from sklearn.metrics.scorer import _check_multimetric_scoring 18 from sklearn.metrics.scorer import _check_multimetric_scoring
26 from sklearn.model_selection import _search, _validation 19 from sklearn.model_selection import _search, _validation
27 from sklearn.model_selection._validation import _score 20 from sklearn.model_selection._validation import _score
28 from sklearn.pipeline import Pipeline 21 from sklearn.pipeline import Pipeline
29 from sklearn.utils import indexable, safe_indexing 22 from sklearn.utils import indexable, safe_indexing
30
31 23
32 _fit_and_score = try_get_attr("galaxy_ml.model_validations", "_fit_and_score") 24 _fit_and_score = try_get_attr("galaxy_ml.model_validations", "_fit_and_score")
33 setattr(_search, "_fit_and_score", _fit_and_score) 25 setattr(_search, "_fit_and_score", _fit_and_score)
34 setattr(_validation, "_fit_and_score", _fit_and_score) 26 setattr(_validation, "_fit_and_score", _fit_and_score)
35 27
54 if swap_value == "": 46 if swap_value == "":
55 continue 47 continue
56 48
57 param_name = p["sp_name"] 49 param_name = p["sp_name"]
58 if param_name.lower().endswith(NON_SEARCHABLE): 50 if param_name.lower().endswith(NON_SEARCHABLE):
59 warnings.warn("Warning: `%s` is not eligible for search and was " "omitted!" % param_name) 51 warnings.warn(
52 "Warning: `%s` is not eligible for search and was "
53 "omitted!" % param_name
54 )
60 continue 55 continue
61 56
62 if not swap_value.startswith(":"): 57 if not swap_value.startswith(":"):
63 safe_eval = SafeEval(load_scipy=True, load_numpy=True) 58 safe_eval = SafeEval(load_scipy=True, load_numpy=True)
64 ev = safe_eval(swap_value) 59 ev = safe_eval(swap_value)
97 groups = kwargs["labels"] 92 groups = kwargs["labels"]
98 n_samples = new_arrays[0].shape[0] 93 n_samples = new_arrays[0].shape[0]
99 index_arr = np.arange(n_samples) 94 index_arr = np.arange(n_samples)
100 test = index_arr[np.isin(groups, group_names)] 95 test = index_arr[np.isin(groups, group_names)]
101 train = index_arr[~np.isin(groups, group_names)] 96 train = index_arr[~np.isin(groups, group_names)]
102 rval = list(chain.from_iterable((safe_indexing(a, train), safe_indexing(a, test)) for a in new_arrays)) 97 rval = list(
98 chain.from_iterable(
99 (safe_indexing(a, train), safe_indexing(a, test)) for a in new_arrays
100 )
101 )
103 else: 102 else:
104 rval = train_test_split(*new_arrays, **kwargs) 103 rval = train_test_split(*new_arrays, **kwargs)
105 104
106 for pos in nones: 105 for pos in nones:
107 rval[pos * 2: 2] = [None, None] 106 rval[pos * 2: 2] = [None, None]
125 if y_true.ndim == 1 or y_true.shape[-1] == 1: 124 if y_true.ndim == 1 or y_true.shape[-1] == 1:
126 pred_probas = pred_probas.ravel() 125 pred_probas = pred_probas.ravel()
127 pred_labels = (pred_probas > 0.5).astype("int32") 126 pred_labels = (pred_probas > 0.5).astype("int32")
128 targets = y_true.ravel().astype("int32") 127 targets = y_true.ravel().astype("int32")
129 if not is_multimetric: 128 if not is_multimetric:
130 preds = pred_labels if scorer.__class__.__name__ == "_PredictScorer" else pred_probas 129 preds = (
130 pred_labels
131 if scorer.__class__.__name__ == "_PredictScorer"
132 else pred_probas
133 )
131 score = scorer._score_func(targets, preds, **scorer._kwargs) 134 score = scorer._score_func(targets, preds, **scorer._kwargs)
132 135
133 return score 136 return score
134 else: 137 else:
135 scores = {} 138 scores = {}
136 for name, one_scorer in scorer.items(): 139 for name, one_scorer in scorer.items():
137 preds = pred_labels if one_scorer.__class__.__name__ == "_PredictScorer" else pred_probas 140 preds = (
141 pred_labels
142 if one_scorer.__class__.__name__ == "_PredictScorer"
143 else pred_probas
144 )
138 score = one_scorer._score_func(targets, preds, **one_scorer._kwargs) 145 score = one_scorer._score_func(targets, preds, **one_scorer._kwargs)
139 scores[name] = score 146 scores[name] = score
140 147
141 # TODO: multi-class metrics 148 # TODO: multi-class metrics
142 # multi-label 149 # multi-label
143 else: 150 else:
144 pred_labels = (pred_probas > 0.5).astype("int32") 151 pred_labels = (pred_probas > 0.5).astype("int32")
145 targets = y_true.astype("int32") 152 targets = y_true.astype("int32")
146 if not is_multimetric: 153 if not is_multimetric:
147 preds = pred_labels if scorer.__class__.__name__ == "_PredictScorer" else pred_probas 154 preds = (
155 pred_labels
156 if scorer.__class__.__name__ == "_PredictScorer"
157 else pred_probas
158 )
148 score, _ = compute_score(preds, targets, scorer._score_func) 159 score, _ = compute_score(preds, targets, scorer._score_func)
149 return score 160 return score
150 else: 161 else:
151 scores = {} 162 scores = {}
152 for name, one_scorer in scorer.items(): 163 for name, one_scorer in scorer.items():
153 preds = pred_labels if one_scorer.__class__.__name__ == "_PredictScorer" else pred_probas 164 preds = (
165 pred_labels
166 if one_scorer.__class__.__name__ == "_PredictScorer"
167 else pred_probas
168 )
154 score, _ = compute_score(preds, targets, one_scorer._score_func) 169 score, _ = compute_score(preds, targets, one_scorer._score_func)
155 scores[name] = score 170 scores[name] = score
156 171
157 return scores 172 return scores
158 173
241 256
242 input_type = params["input_options"]["selected_input"] 257 input_type = params["input_options"]["selected_input"]
243 # tabular input 258 # tabular input
244 if input_type == "tabular": 259 if input_type == "tabular":
245 header = "infer" if params["input_options"]["header1"] else None 260 header = "infer" if params["input_options"]["header1"] else None
246 column_option = params["input_options"]["column_selector_options_1"]["selected_column_selector_option"] 261 column_option = params["input_options"]["column_selector_options_1"][
262 "selected_column_selector_option"
263 ]
247 if column_option in [ 264 if column_option in [
248 "by_index_number", 265 "by_index_number",
249 "all_but_by_index_number", 266 "all_but_by_index_number",
250 "by_header_name", 267 "by_header_name",
251 "all_but_by_header_name", 268 "all_but_by_header_name",
293 n_intervals = sum(1 for line in open(intervals)) 310 n_intervals = sum(1 for line in open(intervals))
294 X = np.arange(n_intervals)[:, np.newaxis] 311 X = np.arange(n_intervals)[:, np.newaxis]
295 312
296 # Get target y 313 # Get target y
297 header = "infer" if params["input_options"]["header2"] else None 314 header = "infer" if params["input_options"]["header2"] else None
298 column_option = params["input_options"]["column_selector_options_2"]["selected_column_selector_option2"] 315 column_option = params["input_options"]["column_selector_options_2"][
316 "selected_column_selector_option2"
317 ]
299 if column_option in [ 318 if column_option in [
300 "by_index_number", 319 "by_index_number",
301 "all_but_by_index_number", 320 "all_but_by_index_number",
302 "by_header_name", 321 "by_header_name",
303 "all_but_by_header_name", 322 "all_but_by_header_name",
311 infile2 = loaded_df[df_key] 330 infile2 = loaded_df[df_key]
312 else: 331 else:
313 infile2 = pd.read_csv(infile2, sep="\t", header=header, parse_dates=True) 332 infile2 = pd.read_csv(infile2, sep="\t", header=header, parse_dates=True)
314 loaded_df[df_key] = infile2 333 loaded_df[df_key] = infile2
315 334
316 y = read_columns(infile2, 335 y = read_columns(
317 c=c, 336 infile2, c=c, c_option=column_option, sep="\t", header=header, parse_dates=True
318 c_option=column_option, 337 )
319 sep='\t',
320 header=header,
321 parse_dates=True)
322 if len(y.shape) == 2 and y.shape[1] == 1: 338 if len(y.shape) == 2 and y.shape[1] == 1:
323 y = y.ravel() 339 y = y.ravel()
324 if input_type == "refseq_and_interval": 340 if input_type == "refseq_and_interval":
325 estimator.set_params(data_batch_generator__features=y.ravel().tolist()) 341 estimator.set_params(data_batch_generator__features=y.ravel().tolist())
326 y = None 342 y = None
327 # end y 343 # end y
328 344
329 # load groups 345 # load groups
330 if groups: 346 if groups:
331 groups_selector = (params["experiment_schemes"]["test_split"]["split_algos"]).pop("groups_selector") 347 groups_selector = (
348 params["experiment_schemes"]["test_split"]["split_algos"]
349 ).pop("groups_selector")
332 350
333 header = "infer" if groups_selector["header_g"] else None 351 header = "infer" if groups_selector["header_g"] else None
334 column_option = groups_selector["column_selector_options_g"]["selected_column_selector_option_g"] 352 column_option = groups_selector["column_selector_options_g"][
353 "selected_column_selector_option_g"
354 ]
335 if column_option in [ 355 if column_option in [
336 "by_index_number", 356 "by_index_number",
337 "all_but_by_index_number", 357 "all_but_by_index_number",
338 "by_header_name", 358 "by_header_name",
339 "all_but_by_header_name", 359 "all_but_by_header_name",
344 364
345 df_key = groups + repr(header) 365 df_key = groups + repr(header)
346 if df_key in loaded_df: 366 if df_key in loaded_df:
347 groups = loaded_df[df_key] 367 groups = loaded_df[df_key]
348 368
349 groups = read_columns(groups, 369 groups = read_columns(
350 c=c, 370 groups,
351 c_option=column_option, 371 c=c,
352 sep='\t', 372 c_option=column_option,
353 header=header, 373 sep="\t",
354 parse_dates=True) 374 header=header,
375 parse_dates=True,
376 )
355 groups = groups.ravel() 377 groups = groups.ravel()
356 378
357 # del loaded_df 379 # del loaded_df
358 del loaded_df 380 del loaded_df
359 381
362 main_est = get_main_estimator(estimator) 384 main_est = get_main_estimator(estimator)
363 if main_est.__class__.__name__ == "IRAPSClassifier": 385 if main_est.__class__.__name__ == "IRAPSClassifier":
364 main_est.set_params(memory=memory) 386 main_est.set_params(memory=memory)
365 387
366 # handle scorer, convert to scorer dict 388 # handle scorer, convert to scorer dict
367 scoring = params['experiment_schemes']['metrics']['scoring'] 389 scoring = params["experiment_schemes"]["metrics"]["scoring"]
368 if scoring is not None: 390 if scoring is not None:
369 # get_scoring() expects secondary_scoring to be a comma separated string (not a list) 391 # get_scoring() expects secondary_scoring to be a comma separated string (not a list)
370 # Check if secondary_scoring is specified 392 # Check if secondary_scoring is specified
371 secondary_scoring = scoring.get("secondary_scoring", None) 393 secondary_scoring = scoring.get("secondary_scoring", None)
372 if secondary_scoring is not None: 394 if secondary_scoring is not None:
383 test_split_options["labels"] = groups 405 test_split_options["labels"] = groups
384 if test_split_options["shuffle"] == "stratified": 406 if test_split_options["shuffle"] == "stratified":
385 if y is not None: 407 if y is not None:
386 test_split_options["labels"] = y 408 test_split_options["labels"] = y
387 else: 409 else:
388 raise ValueError("Stratified shuffle split is not " "applicable on empty target values!") 410 raise ValueError(
411 "Stratified shuffle split is not " "applicable on empty target values!"
412 )
389 413
390 ( 414 (
391 X_train, 415 X_train,
392 X_test, 416 X_test,
393 y_train, 417 y_train,
406 val_split_options["labels"] = groups_train 430 val_split_options["labels"] = groups_train
407 if val_split_options["shuffle"] == "stratified": 431 if val_split_options["shuffle"] == "stratified":
408 if y_train is not None: 432 if y_train is not None:
409 val_split_options["labels"] = y_train 433 val_split_options["labels"] = y_train
410 else: 434 else:
411 raise ValueError("Stratified shuffle split is not " "applicable on empty target values!") 435 raise ValueError(
436 "Stratified shuffle split is not "
437 "applicable on empty target values!"
438 )
412 439
413 ( 440 (
414 X_train, 441 X_train,
415 X_val, 442 X_val,
416 y_train, 443 y_train,
429 estimator.fit(X_train, y_train) 456 estimator.fit(X_train, y_train)
430 457
431 if hasattr(estimator, "evaluate"): 458 if hasattr(estimator, "evaluate"):
432 steps = estimator.prediction_steps 459 steps = estimator.prediction_steps
433 batch_size = estimator.batch_size 460 batch_size = estimator.batch_size
434 generator = estimator.data_generator_.flow(X_test, y=y_test, batch_size=batch_size) 461 generator = estimator.data_generator_.flow(
435 predictions, y_true = _predict_generator(estimator.model_, generator, steps=steps) 462 X_test, y=y_test, batch_size=batch_size
463 )
464 predictions, y_true = _predict_generator(
465 estimator.model_, generator, steps=steps
466 )
436 scores = _evaluate(y_true, predictions, scorer, is_multimetric=True) 467 scores = _evaluate(y_true, predictions, scorer, is_multimetric=True)
437 468
438 else: 469 else:
439 if hasattr(estimator, "predict_proba"): 470 if hasattr(estimator, "predict_proba"):
440 predictions = estimator.predict_proba(X_test) 471 predictions = estimator.predict_proba(X_test)