sonusai 0.18.8__py3-none-any.whl → 0.19.5__py3-none-any.whl

sonusai/genmixdb.py

@@ -76,8 +76,8 @@ generation functions. By default, these are included with the feature data in a
 truth generation is turned on with default settings (see truth section) and a single class, i.e., detecting a single
 type of sound. The truth format is a single float per class representing the probability of activity/presence, and
 multi-class truth is possible by specifying the number of classes and either a scalar index or a vector of indices in
-which to put the truth result. For example, 'num_class: 3' and 'truth_index: 2' adds a 1x3 vector to the feature data
-with truth put in index 2 (others would be 0) for data/target.wav being an audio clip from sound type of class 2.
+which to put the truth result. For example, 'num_class: 3' and 'class_indices: [ 2 ]' adds a 1x3 vector to the feature
+data with truth put in index 2 (others would be 0) for data/target.wav being an audio clip from sound type of class 2.
 
 The mixture is created with potential data augmentation functions in the following way:
 1. apply noise augmentation rule
@@ -112,6 +112,7 @@ targets:
 will find all .wav files in the specified directories and process them as targets.
 
 """
+
 import signal
 from dataclasses import dataclass
 
@@ -124,7 +125,7 @@ def signal_handler(_sig, _frame):
 
     from sonusai import logger
 
-    logger.info('Canceled due to keyboard interrupt')
+    logger.info("Canceled due to keyboard interrupt")
     sys.exit(1)
 
 
@@ -133,34 +134,34 @@ signal.signal(signal.SIGINT, signal_handler)
 
 @dataclass
 class MPGlobal:
-    mixdb: MixtureDatabase = None
-    save_mix: bool = None
-    save_ft: bool = None
-    save_segsnr: bool = None
+    mixdb: MixtureDatabase
+    save_mix: bool
+    save_ft: bool
+    save_segsnr: bool
 
 
-MP_GLOBAL = MPGlobal()
+MP_GLOBAL: MPGlobal
 
 
-def genmixdb(location: str,
-             save_mix: bool = False,
-             save_ft: bool = False,
-             save_segsnr: bool = False,
-             logging: bool = True,
-             show_progress: bool = False,
-             test: bool = False,
-             save_json: bool = False) -> MixtureDatabase:
+def genmixdb(
+    location: str,
+    save_mix: bool = False,
+    save_ft: bool = False,
+    save_segsnr: bool = False,
+    logging: bool = True,
+    show_progress: bool = False,
+    test: bool = False,
+    save_json: bool = False,
+) -> MixtureDatabase:
     from random import seed
 
     import yaml
-    from tqdm import tqdm
 
-    from sonusai import SonusAIError
     from sonusai import logger
-    from sonusai.mixture import AugmentationRule
-    from sonusai.mixture import MixtureDatabase
     from sonusai.mixture import SAMPLE_BYTES
     from sonusai.mixture import SAMPLE_RATE
+    from sonusai.mixture import AugmentationRule
+    from sonusai.mixture import MixtureDatabase
     from sonusai.mixture import balance_targets
     from sonusai.mixture import generate_mixtures
     from sonusai.mixture import get_all_snrs_from_config
@@ -182,11 +183,13 @@ def genmixdb(location: str,
     from sonusai.mixture import populate_spectral_mask_table
     from sonusai.mixture import populate_target_file_table
     from sonusai.mixture import populate_top_table
+    from sonusai.mixture import populate_truth_parameters_table
     from sonusai.mixture import update_mixid_width
     from sonusai.utils import dataclass_from_dict
     from sonusai.utils import human_readable_size
-    from sonusai.utils import pp_tqdm_imap
+    from sonusai.utils import par_track
     from sonusai.utils import seconds_to_hms
+    from sonusai.utils import track
 
     config = load_config(location)
     initialize_db(location=location, test=test)
@@ -197,113 +200,116 @@ def genmixdb(location: str,
     populate_class_label_table(location, config, test)
     populate_class_weights_threshold_table(location, config, test)
     populate_spectral_mask_table(location, config, test)
+    populate_truth_parameters_table(location, config, test)
 
-    seed(config['seed'])
+    seed(config["seed"])
 
     if logging:
-        logger.debug(f'Seed: {config["seed"]}')
-        logger.debug('Configuration:')
+        logger.debug(f"Seed: {config['seed']}")
+        logger.debug("Configuration:")
         logger.debug(yaml.dump(config))
 
     if logging:
-        logger.info('Collecting targets')
+        logger.info("Collecting targets")
 
     target_files = get_target_files(config, show_progress=show_progress)
 
     if len(target_files) == 0:
-        raise SonusAIError('Canceled due to no targets')
+        raise RuntimeError("Canceled due to no targets")
 
     populate_target_file_table(location, target_files, test)
 
     if logging:
-        logger.debug('List of targets:')
+        logger.debug("List of targets:")
         logger.debug(yaml.dump([target.name for target in mixdb.target_files], default_flow_style=False))
-        logger.debug('')
+        logger.debug("")
 
     if logging:
-        logger.info('Collecting noises')
+        logger.info("Collecting noises")
 
     noise_files = get_noise_files(config, show_progress=show_progress)
 
     populate_noise_file_table(location, noise_files, test)
 
     if logging:
-        logger.debug('List of noises:')
+        logger.debug("List of noises:")
         logger.debug(yaml.dump([noise.name for noise in mixdb.noise_files], default_flow_style=False))
-        logger.debug('')
+        logger.debug("")
 
     if logging:
-        logger.info('Collecting impulse responses')
+        logger.info("Collecting impulse responses")
 
     impulse_response_files = get_impulse_response_files(config)
 
     populate_impulse_response_file_table(location, impulse_response_files, test)
 
     if logging:
-        logger.debug('List of impulse responses:')
+        logger.debug("List of impulse responses:")
         logger.debug(
-            yaml.dump([impulse_response for impulse_response in mixdb.impulse_response_files],
-                      default_flow_style=False))
-        logger.debug('')
+            yaml.dump(
+                [entry.file for entry in mixdb.impulse_response_files],
+                default_flow_style=False,
+            )
+        )
+        logger.debug("")
 
     if logging:
-        logger.info('Collecting target augmentations')
+        logger.info("Collecting target augmentations")
 
-    target_augmentations = get_augmentation_rules(rules=config['target_augmentations'],
-                                                  num_ir=mixdb.num_impulse_response_files)
+    target_augmentations = get_augmentation_rules(
+        rules=config["target_augmentations"], num_ir=mixdb.num_impulse_response_files
+    )
     mixups = get_mixups(target_augmentations)
 
     if logging:
         for mixup in mixups:
-            logger.debug(f'Expanded list of target augmentation rules for mixup of {mixup}:')
+            logger.debug(f"Expanded list of target augmentation rules for mixup of {mixup}:")
             for target_augmentation in get_target_augmentations_for_mixup(target_augmentations, mixup):
                 ta_dict = target_augmentation.to_dict()
-                del ta_dict['mixup']
-                logger.debug(f'- {ta_dict}')
-            logger.debug('')
+                del ta_dict["mixup"]
+                logger.debug(f"- {ta_dict}")
+            logger.debug("")
 
     if logging:
-        logger.info('Collecting noise augmentations')
+        logger.info("Collecting noise augmentations")
 
-    noise_augmentations = get_augmentation_rules(rules=config['noise_augmentations'],
-                                                 num_ir=mixdb.num_impulse_response_files)
+    noise_augmentations = get_augmentation_rules(
+        rules=config["noise_augmentations"], num_ir=mixdb.num_impulse_response_files
+    )
 
     if logging:
-        logger.debug('Expanded list of noise augmentations:')
+        logger.debug("Expanded list of noise augmentations:")
         for noise_augmentation in noise_augmentations:
             na_dict = noise_augmentation.to_dict()
-            del na_dict['mixup']
-            logger.debug(f'- {na_dict}')
-        logger.debug('')
+            del na_dict["mixup"]
+            logger.debug(f"- {na_dict}")
+        logger.debug("")
 
     if logging:
-        logger.debug(f'SNRs: {config["snrs"]}\n')
-        logger.debug(f'Random SNRs: {config["random_snrs"]}\n')
-        logger.debug(f'Noise mix mode: {mixdb.noise_mix_mode}\n')
-        logger.debug(f'Spectral masks:')
+        logger.debug(f"SNRs: {config['snrs']}\n")
+        logger.debug(f"Random SNRs: {config['random_snrs']}\n")
+        logger.debug(f"Noise mix mode: {mixdb.noise_mix_mode}\n")
+        logger.debug("Spectral masks:")
         for spectral_mask in mixdb.spectral_masks:
-            logger.debug(f'- {spectral_mask}')
-        logger.debug('')
-
-    if mixdb.truth_mutex and any(mixup > 1 for mixup in mixups):
-        raise SonusAIError(f'Mutex truth mode is not compatible with mixup')
+            logger.debug(f"- {spectral_mask}")
+        logger.debug("")
 
     if logging:
-        logger.info('Collecting augmented targets')
+        logger.info("Collecting augmented targets")
 
     augmented_targets = get_augmented_targets(target_files, target_augmentations, mixups)
 
-    if config['class_balancing']:
-        class_balancing_augmentation = dataclass_from_dict(AugmentationRule, config['class_balancing_augmentation'])
+    if config["class_balancing"]:
+        class_balancing_augmentation = dataclass_from_dict(AugmentationRule, config["class_balancing_augmentation"])
         augmented_targets, target_augmentations = balance_targets(
             augmented_targets=augmented_targets,
             targets=target_files,
             target_augmentations=target_augmentations,
             class_balancing_augmentation=class_balancing_augmentation,
             num_classes=mixdb.num_classes,
-            truth_mutex=mixdb.truth_mutex,
             num_ir=mixdb.num_impulse_response_files,
-            mixups=mixups)
+            mixups=mixups,
+        )
 
     target_audio_samples = sum([targets.samples for targets in mixdb.target_files])
     target_audio_duration = target_audio_samples / SAMPLE_RATE
@@ -311,13 +317,17 @@ def genmixdb(location: str,
     noise_audio_samples = noise_audio_duration * SAMPLE_RATE
 
     if logging:
-        logger.info('')
-        logger.info(f'Target audio: {mixdb.num_target_files} files, '
-                    f'{human_readable_size(target_audio_samples * SAMPLE_BYTES, 1)}, '
-                    f'{seconds_to_hms(seconds=target_audio_duration)}')
-        logger.info(f'Noise audio: {mixdb.num_noise_files} files, '
-                    f'{human_readable_size(noise_audio_samples * SAMPLE_BYTES, 1)}, '
-                    f'{seconds_to_hms(seconds=noise_audio_duration)}')
+        logger.info("")
+        logger.info(
+            f"Target audio: {mixdb.num_target_files} files, "
+            f"{human_readable_size(target_audio_samples * SAMPLE_BYTES, 1)}, "
+            f"{seconds_to_hms(seconds=target_audio_duration)}"
+        )
+        logger.info(
+            f"Noise audio: {mixdb.num_noise_files} files, "
+            f"{human_readable_size(noise_audio_samples * SAMPLE_BYTES, 1)}, "
+            f"{seconds_to_hms(seconds=noise_audio_duration)}"
+        )
 
     used_noise_files, used_noise_samples, mixtures = generate_mixtures(
         noise_mix_mode=mixdb.noise_mix_mode,
@@ -330,41 +340,48 @@ def genmixdb(location: str,
         all_snrs=get_all_snrs_from_config(config),
         mixups=mixups,
         num_classes=mixdb.num_classes,
-        truth_mutex=mixdb.truth_mutex,
         feature_step_samples=mixdb.feature_step_samples,
-        num_ir=mixdb.num_impulse_response_files)
+        num_ir=mixdb.num_impulse_response_files,
+    )
 
     num_mixtures = len(mixtures)
     update_mixid_width(location, num_mixtures, test)
 
     if logging:
-        logger.info('')
-        logger.info(f'Found {num_mixtures:,} mixtures to process')
+        logger.info("")
+        logger.info(f"Found {num_mixtures:,} mixtures to process")
 
     total_duration = float(sum([mixture.samples for mixture in mixtures])) / SAMPLE_RATE
 
     if logging:
-        log_duration_and_sizes(total_duration=total_duration,
-                               num_classes=mixdb.num_classes,
-                               feature_step_samples=mixdb.feature_step_samples,
-                               feature_parameters=mixdb.feature_parameters,
-                               stride=mixdb.fg_stride,
-                               desc='Estimated')
-        logger.info(f'Feature shape:        '
-                    f'{mixdb.fg_stride} x {mixdb.feature_parameters} '
-                    f'({mixdb.fg_stride * mixdb.feature_parameters} total params)')
-        logger.info(f'Feature samples:      {mixdb.feature_samples} samples ({mixdb.feature_ms} ms)')
-        logger.info(f'Feature step samples: {mixdb.feature_step_samples} samples ({mixdb.feature_step_ms} ms)')
-        logger.info('')
+        log_duration_and_sizes(
+            total_duration=total_duration,
+            num_classes=mixdb.num_classes,
+            feature_step_samples=mixdb.feature_step_samples,
+            feature_parameters=mixdb.feature_parameters,
+            stride=mixdb.fg_stride,
+            desc="Estimated",
+        )
+        logger.info(
+            f"Feature shape:        "
+            f"{mixdb.fg_stride} x {mixdb.feature_parameters} "
+            f"({mixdb.fg_stride * mixdb.feature_parameters} total params)"
+        )
+        logger.info(f"Feature samples:      {mixdb.feature_samples} samples ({mixdb.feature_ms} ms)")
+        logger.info(f"Feature step samples: {mixdb.feature_step_samples} samples ({mixdb.feature_step_ms} ms)")
+        logger.info("")
 
     # Fill in the details
     if logging:
-        logger.info('Generating mixtures')
-    progress = tqdm(total=num_mixtures, disable=not show_progress)
-    mixtures = pp_tqdm_imap(_process_mixture, mixtures,
-                            progress=progress,
-                            initializer=_initializer,
-                            initargs=(location, save_mix, save_ft, save_segsnr, test))
+        logger.info("Generating mixtures")
+    progress = track(total=num_mixtures, disable=not show_progress)
+    mixtures = par_track(
+        _process_mixture,
+        mixtures,
+        progress=progress,
+        initializer=_initializer,
+        initargs=(location, save_mix, save_ft, save_segsnr, test),
+    )
     progress.close()
 
     populate_mixture_table(location, mixtures, test)
@@ -378,20 +395,22 @@ def genmixdb(location: str,
     noise_samples_percent = (float(used_noise_samples) / float(noise_audio_samples)) * 100
 
     if logging:
-        log_duration_and_sizes(total_duration=total_duration,
-                               num_classes=mixdb.num_classes,
-                               feature_step_samples=mixdb.feature_step_samples,
-                               feature_parameters=mixdb.feature_parameters,
-                               stride=mixdb.fg_stride,
-                               desc='Actual')
-        logger.info('')
-        logger.info(f'Used {noise_files_percent:,.0f}% of noise files')
-        logger.info(f'Used {noise_samples_percent:,.0f}% of noise audio')
-        logger.info('')
+        log_duration_and_sizes(
+            total_duration=total_duration,
+            num_classes=mixdb.num_classes,
+            feature_step_samples=mixdb.feature_step_samples,
+            feature_parameters=mixdb.feature_parameters,
+            stride=mixdb.fg_stride,
+            desc="Actual",
+        )
+        logger.info("")
+        logger.info(f"Used {noise_files_percent:,.0f}% of noise files")
+        logger.info(f"Used {noise_samples_percent:,.0f}% of noise audio")
+        logger.info("")
 
     if not test and save_json:
         if logging:
-            logger.info(f'Writing JSON version of database to {location}')
+            logger.info(f"Writing JSON version of database to {location}")
         mixdb = MixtureDatabase(location)
         mixdb.save()
 
@@ -399,10 +418,14 @@ def genmixdb(location: str,
 
 
 def _initializer(location: str, save_mix: bool, save_ft: bool, save_segsnr: bool, test: bool) -> None:
-    MP_GLOBAL.mixdb = MixtureDatabase(location, test)
-    MP_GLOBAL.save_mix = save_mix
-    MP_GLOBAL.save_ft = save_ft
-    MP_GLOBAL.save_segsnr = save_segsnr
+    global MP_GLOBAL
+
+    MP_GLOBAL = MPGlobal(
+        mixdb=MixtureDatabase(location, test),
+        save_mix=save_mix,
+        save_ft=save_ft,
+        save_segsnr=save_segsnr,
+    )
 
 
 def _process_mixture(mixture: Mixture) -> Mixture:
@@ -410,11 +433,13 @@ def _process_mixture(mixture: Mixture) -> Mixture:
 
     from sonusai.mixture import get_ft
     from sonusai.mixture import get_segsnr
-    from sonusai.mixture import get_truth_t
+    from sonusai.mixture import get_truth
     from sonusai.mixture import update_mixture
-    from sonusai.mixture import write_mixture_data
+    from sonusai.mixture import write_cached_data
     from sonusai.mixture import write_mixture_metadata
 
+    global MP_GLOBAL
+
     with_data = MP_GLOBAL.save_mix or MP_GLOBAL.save_ft
     mixdb = MP_GLOBAL.mixdb
 
@@ -424,31 +449,41 @@ def _process_mixture(mixture: Mixture) -> Mixture:
         write_data: list[tuple[str, Any]] = []
 
         if MP_GLOBAL.save_mix:
-            write_data.append(('targets', genmix_data.targets))
-            write_data.append(('noise', genmix_data.noise))
-            write_data.append(('mixture', genmix_data.mixture))
+            write_data.append(("targets", genmix_data.targets))
+            write_data.append(("noise", genmix_data.noise))
+            write_data.append(("mixture", genmix_data.mixture))
 
         if MP_GLOBAL.save_ft:
-            truth_t = get_truth_t(mixdb=mixdb,
-                                  mixture=mixture,
-                                  targets_audio=genmix_data.targets,
-                                  noise_audio=genmix_data.noise,
-                                  mixture_audio=genmix_data.mixture)
-            feature, truth_f = get_ft(mixdb=mixdb,
-                                      mixture=mixture,
-                                      mixture_audio=genmix_data.mixture,
-                                      truth_t=truth_t)
-            write_data.append(('feature', feature))
-            write_data.append(('truth_f', truth_f))
+            if genmix_data.targets is None or genmix_data.noise is None or genmix_data.mixture is None:
+                raise RuntimeError("Mixture data was not generated properly")
+            truth_t = get_truth(
+                mixdb=mixdb,
+                mixture=mixture,
+                targets_audio=genmix_data.targets,
+                noise_audio=genmix_data.noise,
+                mixture_audio=genmix_data.mixture,
+            )
+            feature, truth_f = get_ft(
+                mixdb=mixdb,
+                mixture=mixture,
+                mixture_audio=genmix_data.mixture,
+                truth_t=truth_t,
+            )
+            write_data.append(("feature", feature))
+            write_data.append(("truth_f", truth_f))
 
             if MP_GLOBAL.save_segsnr:
-                segsnr = get_segsnr(mixdb=mixdb,
-                                    mixture=mixture,
-                                    target_audio=genmix_data.target,
-                                    noise=genmix_data.noise)
-                write_data.append(('segsnr', segsnr))
-
-        write_mixture_data(mixdb, mixture, write_data)
+                if genmix_data.target is None:
+                    raise RuntimeError("Target data was not generated properly")
+                segsnr = get_segsnr(
+                    mixdb=mixdb,
+                    mixture=mixture,
+                    target_audio=genmix_data.target,
+                    noise=genmix_data.noise,
+                )
+                write_data.append(("segsnr", segsnr))
+
+        write_cached_data(mixdb.location, "mixture", mixture.name, write_data)
         write_mixture_metadata(mixdb, mixture)
 
     return mixture
@@ -478,13 +513,13 @@ def main() -> None:
     from sonusai.mixture import load_config
     from sonusai.utils import seconds_to_hms
 
-    verbose = args['--verbose']
-    save_mix = args['--mix']
-    save_ft = args['--ft']
-    save_segsnr = args['--segsnr']
-    dryrun = args['--dryrun']
-    save_json = args['--json']
-    location = args['LOC']
+    verbose = args["--verbose"]
+    save_mix = args["--mix"]
+    save_ft = args["--ft"]
+    save_segsnr = args["--segsnr"]
+    dryrun = args["--dryrun"]
+    save_json = args["--json"]
+    location = args["LOC"]
 
     start_time = time.monotonic()
 
@@ -493,30 +528,36 @@ def main() -> None:
 
     makedirs(location, exist_ok=True)
 
-    create_file_handler(join(location, 'genmixdb.log'))
+    create_file_handler(join(location, "genmixdb.log"))
     update_console_handler(verbose)
-    initial_log_messages('genmixdb')
+    initial_log_messages("genmixdb")
 
     if dryrun:
         config = load_config(location)
-        logger.info('Dryrun configuration:')
+        logger.info("Dryrun configuration:")
         logger.info(yaml.dump(config))
         return
 
-    logger.info(f'Creating mixture database for {location}')
-    logger.info('')
-
-    genmixdb(location=location,
-             save_mix=save_mix,
-             save_ft=save_ft,
-             save_segsnr=save_segsnr,
-             show_progress=True,
-             save_json=save_json)
+    logger.info(f"Creating mixture database for {location}")
+    logger.info("")
+
+    try:
+        genmixdb(
+            location=location,
+            save_mix=save_mix,
+            save_ft=save_ft,
+            save_segsnr=save_segsnr,
+            show_progress=True,
+            save_json=save_json,
+        )
+    except Exception as e:
+        logger.debug(e)
+        raise
 
     end_time = time.monotonic()
-    logger.info(f'Completed in {seconds_to_hms(seconds=end_time - start_time)}')
-    logger.info('')
+    logger.info(f"Completed in {seconds_to_hms(seconds=end_time - start_time)}")
+    logger.info("")
 
 
-if __name__ == '__main__':
+if __name__ == "__main__":
     main()