Added file used for development

nns/Cnn_test.py

@@ -129,7 +129,10 @@ class NeuralNetwork(object):
     def predict(self, x, clean_signal=True, return_probs=False):
         """Given sequences input as x, predict if they contain target k-mer.
         Assumes the sequence x is a read that has been normalised,
-        but not cut into smaller chunks
+        but not cut into smaller chunks.
+
+        Function is mainly written to be called from train_nn.py.
+        Not for final inference.
 
         :param x: Squiggle as numeric representation
         :type x: np.ndarray

trash_debug_CNN.py

+import h5py
+from db_building.TrainingRead import Read, TrainingRead
+from pathlib import Path
+from inference.InferenceModel import InferenceModel
+from nns.Cnn_test import NeuralNetwork
+import tensorflow.keras.backend as K
+import numpy as np
+import tensorflow as tf
+import time
+
+
+from nns.keras_metrics_from_logits import precision, recall, binary_accuracy
+
+def main():
+    # Hunt for positive read
+    input_length = 250
+    stride = input_length // 2
+    target_kmer = 'AGGAGAGT'
+
+    # for file in Path('/home/noord087/lustre_link/HoiCarlos/16Sreads_mockcommunity/demultiplexed_reads/files_for_initial_training/test').iterdir():
+    #     print(f'Scanning {file}')
+    #     with h5py.File(file, 'r') as h5_file:
+    #         try:
+    #             train_read = TrainingRead(h5_file, 'median',
+    #                                           'Analyses/RawGenomeCorrected_000', 8)
+    #             if [i for i in train_read.condensed_events
+    #                 if i[0] == target_kmer]:
+    #                 print(f"found in {file}")
+    #                 break
+    #         except KeyError as e:
+    #             print('Got keyerror, continuing')
+    #
+    # return
+
+    pos_read_path = Path(
+        '/home/noord087/lustre_link/HoiCarlos/16Sreads_mockcommunity/demultiplexed_reads/files_for_initial_training/test/L0144169_20181212_FAK22428_MN19628_sequencing_run_16Srhizhome_2_99947_read_129882_ch_413_strand.fast5'
+    )
+
+    neg_read_path = Path(
+        '/home/noord087/lustre_link/HoiCarlos/16Sreads_mockcommunity/demultiplexed_reads/files_for_initial_training/test/L0144169_20181212_FAK22428_MN19628_sequencing_run_16Srhizosphere_1_66037_read_11482_ch_104_strand.fast5')
+    with h5py.File(pos_read_path, 'r') as f:
+        pos_read = Read(f, 'median')
+
+    with h5py.File(neg_read_path, 'r') as f:
+        neg_read = Read(f, 'median')
+
+    split_pos_read = pos_read.get_split_raw_read(input_length, stride=input_length)
+    split_neg_read = neg_read.get_split_raw_read(input_length, stride=input_length)
+
+    def loss_fun(y_true,
+                 y_pred):  # just dummy to satisfy the stupid thing being there
+        msk = np.zeros(100, dtype=bool)
+        msk[50] = True
+        y_pred_single = tf.boolean_mask(y_pred, msk, axis=1)
+        return K.binary_crossentropy(K.cast(y_true, K.floatx()), y_pred_single,
+                                     from_logits=True)
+
+    cnn = tf.keras.models.load_model(
+        '/lustre/BIF/nobackup/noord087/mscthesis/baseless/baseless_250_width_uncentered_kmer/nns/AGGAGAGT/nn.h5',
+        custom_objects={'precision': precision,
+                        'recall': recall,
+                        'binary_accuracy': binary_accuracy,
+                        'loss_fun': loss_fun})
+
+    start = time.time()
+    pos_posteriors = cnn.predict(split_pos_read)
+    neg_posteriors = cnn.predict(split_neg_read)
+    print(f'.predict method: {time.time() - start} seconds')
+
+    start = time.time()
+    pos_posteriors = cnn(split_pos_read)
+    neg_posteriors = cnn(split_neg_read)
+    print(f'Direct call: {time.time() - start} seconds')
+
+    print(f'{np.sum(pos_posteriors > 0)=}')
+    print(f'{np.sum(neg_posteriors > 0)=}')
+
+
+if __name__ == '__main__':
+    main()