1""" 2/* Copyright (c) 2023 Amazon 3 Written by Jan Buethe */ 4/* 5 Redistribution and use in source and binary forms, with or without 6 modification, are permitted provided that the following conditions 7 are met: 8 9 - Redistributions of source code must retain the above copyright 10 notice, this list of conditions and the following disclaimer. 11 12 - Redistributions in binary form must reproduce the above copyright 13 notice, this list of conditions and the following disclaimer in the 14 documentation and/or other materials provided with the distribution. 15 16 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 17 ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 18 LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 19 A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER 20 OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, 21 EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, 22 PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR 23 PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF 24 LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING 25 NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS 26 SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 27*/ 28""" 29 30import argparse 31import os 32from uuid import UUID 33from collections import OrderedDict 34import pickle 35 36 37import torch 38import numpy as np 39 40import utils 41 42 43 44parser = argparse.ArgumentParser() 45parser.add_argument("input", type=str, help="input folder containing multi-run output") 46parser.add_argument("tag", type=str, help="tag for multi-run experiment") 47parser.add_argument("csv", type=str, help="name for output csv") 48 49 50def is_uuid(val): 51 try: 52 UUID(val) 53 return True 54 except: 55 return False 56 57 58def collect_results(folder): 59 60 training_folder = os.path.join(folder, 'training') 61 testing_folder = os.path.join(folder, 'testing') 62 63 # validation loss 64 checkpoint = torch.load(os.path.join(training_folder, 'checkpoints', 'checkpoint_finalize_epoch_1.pth'), map_location='cpu') 65 validation_loss = checkpoint['validation_loss'] 66 67 # eval_warpq 68 eval_warpq = utils.data.parse_warpq_scores(os.path.join(training_folder, 'out_finalize.txt'))[-1] 69 70 # testing results 71 testing_results = utils.data.collect_test_stats(os.path.join(testing_folder, 'final')) 72 73 results = OrderedDict() 74 results['eval_loss'] = validation_loss 75 results['eval_warpq'] = eval_warpq 76 results['pesq_mean'] = testing_results['pesq'][0] 77 results['warpq_mean'] = testing_results['warpq'][0] 78 results['pitch_error_mean'] = testing_results['pitch_error'][0] 79 results['voicing_error_mean'] = testing_results['voicing_error'][0] 80 81 return results 82 83def print_csv(path, results, tag, ranks=None, header=True): 84 85 metrics = next(iter(results.values())).keys() 86 if ranks is not None: 87 rank_keys = next(iter(ranks.values())).keys() 88 else: 89 rank_keys = [] 90 91 with open(path, 'w') as f: 92 if header: 93 f.write("uuid, tag") 94 95 for metric in metrics: 96 f.write(f", {metric}") 97 98 for rank in rank_keys: 99 f.write(f", {rank}") 100 101 f.write("\n") 102 103 104 for uuid, values in results.items(): 105 f.write(f"{uuid}, {tag}") 106 107 for val in values.values(): 108 f.write(f", {val:10.8f}") 109 110 for rank in rank_keys: 111 f.write(f", {ranks[uuid][rank]:4d}") 112 113 f.write("\n") 114 115def get_ranks(results): 116 117 metrics = list(next(iter(results.values())).keys()) 118 119 positive = {'pesq_mean', 'mix'} 120 121 ranks = OrderedDict() 122 for key in results.keys(): 123 ranks[key] = OrderedDict() 124 125 for metric in metrics: 126 sign = -1 if metric in positive else 1 127 128 x = sorted([(key, value[metric]) for key, value in results.items()], key=lambda x: sign * x[1]) 129 x = [y[0] for y in x] 130 131 for key in results.keys(): 132 ranks[key]['rank_' + metric] = x.index(key) + 1 133 134 return ranks 135 136def analyse_metrics(results): 137 metrics = ['eval_loss', 'pesq_mean', 'warpq_mean', 'pitch_error_mean', 'voicing_error_mean'] 138 139 x = [] 140 for metric in metrics: 141 x.append([val[metric] for val in results.values()]) 142 143 x = np.array(x) 144 145 print(x) 146 147def add_mix_metric(results): 148 metrics = ['eval_loss', 'pesq_mean', 'warpq_mean', 'pitch_error_mean', 'voicing_error_mean'] 149 150 x = [] 151 for metric in metrics: 152 x.append([val[metric] for val in results.values()]) 153 154 x = np.array(x).transpose() * np.array([-1, 1, -1, -1, -1]) 155 156 z = (x - np.mean(x, axis=0)) / np.std(x, axis=0) 157 158 print(f"covariance matrix for normalized scores of {metrics}:") 159 print(np.cov(z.transpose())) 160 161 score = np.mean(z, axis=1) 162 163 for i, key in enumerate(results.keys()): 164 results[key]['mix'] = score[i].item() 165 166if __name__ == "__main__": 167 args = parser.parse_args() 168 169 uuids = sorted([x for x in os.listdir(args.input) if os.path.isdir(os.path.join(args.input, x)) and is_uuid(x)]) 170 171 172 results = OrderedDict() 173 174 for uuid in uuids: 175 results[uuid] = collect_results(os.path.join(args.input, uuid)) 176 177 178 add_mix_metric(results) 179 180 ranks = get_ranks(results) 181 182 183 184 csv = args.csv if args.csv.endswith('.csv') else args.csv + '.csv' 185 186 print_csv(args.csv, results, args.tag, ranks=ranks) 187 188 189 with open(csv[:-4] + '.pickle', 'wb') as f: 190 pickle.dump(results, f, protocol=pickle.HIGHEST_PROTOCOL)