sonusai 1.0.16__cp311-abi3-macosx_10_12_x86_64.whl

sonusai/deprecated/tplot.py

@@ -0,0 +1,365 @@
+# mypy: ignore-errors
+"""sonusai tplot
+
+usage: tplot [-hv] [-r ROOT] [-c CONFIG] [-t TARGET] [-m MODEL] [-f FEATURE] [--sedthr SEDTHR]
+
+options:
+   -h, --help
+   -v, --verbose            Be verbose.
+   -r, --root ROOT          Root of a gentcst tree.
+   -c, --config CONFIG      YAML config file.
+   -t, --target TARGET      Target .wav or .txt, else use target section in CONFIG.
+   -m, --model MODEL        Aaware ONNX model file.
+   -f, --feature FEATURE    Feature override, e.g., gfh64to128.
+   --sedthr SEDTHR          SED thresholds override, e.g., '[-24, -27, -30]'.
+
+   CONFIG is used to override the default configuration, thus it does not need to be a complete
+   genmixdb config parameter set, only desired overrides like truth function and truth config.
+
+   A TARGET .txt file supports the same format as SonusAI genmixdb which is a list of regexes
+   for finding sound files.
+
+   ROOT, CONFIG, and TARGET are optional, but at least one of them must be specified.
+   Behavior is as follows in order of precedence:
+
+    1)  If ROOT is provided, use the given directory as the root directory of
+        a gentcst tree. The config will be generated from the top down to the current leaf.
+    2)  If CONFIG is provided, use the given YAML file to override the config.
+    3)  If TARGET is provided, use the given .wav or .txt file to override the target
+        section of the config.
+    4)  If FEATURE and/or SEDTHR are provided, use them to override the config.
+
+   If TARGET is given without a CONFIG, look for a corresponding YAML file to use as a
+   config. For example,
+
+        sonusai tplot -t 4-esc50-dogbark.txt
+
+   will first look for '4-esc50-dogbark.yml' and then for 'config.yml'. If neither
+   candidate was found and ROOT was also not specified then give an error. If
+   ROOT was specified then use the hierarchical config.
+
+   A multi-page plot TARGET-tplot.pdf or CONFIG-tplot.pdf is generated.
+
+"""
+
+import signal
+
+
+def signal_handler(_sig, _frame):
+    import sys
+
+    from sonusai import logger
+
+    logger.info("Canceled due to keyboard interrupt")
+    sys.exit(1)
+
+
+signal.signal(signal.SIGINT, signal_handler)
+
+
+# TODO: re-work for modern mixdb API
+def tplot(
+    root: str | None = None,
+    config_file: str | None = None,
+    target_file: str | None = None,
+    model_file: str | None = None,
+    feature: str | None = None,
+    thresholds: str | None = None,
+    verbose: bool = False,
+) -> None:
+    import json
+    import os
+
+    import matplotlib.pyplot as plt
+    import numpy as np
+    import yaml
+    from matplotlib.backends.backend_pdf import PdfPages
+    from pyaaware import Predict
+
+    import sonusai
+    from sonusai import SonusAIError
+    from sonusai import logger
+    from sonusai.genft import genft
+    from sonusai.genmix import genmix
+    from sonusai.genmixdb import genmixdb
+    from sonusai.mixture import get_default_config
+    from sonusai.mixture import update_config_from_file
+    from sonusai.mixture import update_config_from_hierarchy
+
+    figsize = (11, 8.5)
+
+    if config_file is None and target_file is None and root is None:
+        raise SonusAIError("Error: No config provided.")
+
+    target_path = ""
+    config = get_default_config()
+
+    if root is not None:
+        leaf = os.getcwd()
+        target_path = os.path.basename(leaf)
+        config = update_config_from_hierarchy(root=root, leaf=leaf, config=config)
+        logger.debug(f"Hierarchical config for {leaf} using {root} as the root")
+
+    if config_file is not None:
+        target_path = config_file
+        config = update_config_from_file(config_file, config)
+        logger.debug(f"Config from {config_file}")
+
+    if target_file is not None:
+        target_path = target_file
+        if not os.path.exists(target_file):
+            raise SonusAIError(f"{target_file} does not exist.")
+
+        config["targets"] = [{"name": target_file}]
+        logger.debug(f"Targets from {target_file}")
+
+        if config_file is None:
+            target_config = f"{os.path.splitext(target_file)[0]}.yml"
+            if os.path.exists(target_config):
+                config = update_config_from_file(target_config, config)
+                logger.debug(f"Config from {target_config}")
+            else:
+                target_config = "config.yml"
+                if os.path.exists(target_config):
+                    config = update_config_from_file(target_config, config)
+                    logger.debug(f"Config from {target_config}")
+                elif root is None:
+                    raise SonusAIError(f"Error: Could not determine config for {target_file}.")
+
+    file_base = os.path.splitext(target_path)[0]
+    file_name = os.path.split(file_base)[1]
+    output = f"./{file_name}-tplot.pdf"
+
+    if len(config["truth_settings"]) > 1:
+        raise SonusAIError("Number of truth_settings is more than one. This is not supported yet.")
+
+    if thresholds is not None:
+        thresholds = json.loads(thresholds)
+        if thresholds != config["truth_settings"][0]["config"]["thresholds"]:
+            config["truth_settings"][0]["config"]["thresholds"] = thresholds
+            logger.debug(f"Override SED thresholds with {thresholds}")
+
+    if feature is not None and feature != config["feature"]:
+        config["feature"] = feature
+        logger.debug(f"Override feature with {feature}")
+
+    logger.debug("")
+
+    # Run genmixdb, genmix, and genft
+    mixdb = genmixdb(config=config, logging=verbose)
+    num_mixtures = mixdb.num_mixtures  # Number of mixtures
+    logger.info(f"Generating data for {num_mixtures} mixture plots")
+    audio_data = genmix(
+        mixdb=mixdb,
+        compute_segsnr=False,
+        compute_truth=False,
+    )
+    feature_data = genft(mixdb=mixdb, compute_segsnr=False)
+
+    # If model provided, read and check it
+    predict_data = np.zeros(0)
+    if model_file is not None:
+        import onnx
+
+        # model is an in-memory ModelProto
+        model = onnx.load(model_file)
+        if len(model.metadata_props) < 5:
+            logger.warn("Model metadata indicates this is not an Aaware model, ignoring.")
+        else:
+            if model.metadata_props[4].key != "feature":
+                logger.warn("Model metadata does not have Aaware feature, ignoring.")
+            else:
+                model_feature = model.metadata_props[4].value
+                logger.debug(f"Model feature is {model_feature}")
+                # TODO Check and read other params, flatten, addch, etc.
+                logger.info(f"Running prediction with {model_file}")
+                predict = Predict(model_file)
+                predict_data = predict.execute(feature_data.feature)
+
+    num_samples = audio_data.mixture.shape[0]  # Total number of samples over all mixtures
+    num_features = feature_data.feature.shape[0]  # Total number of feature frames over all mixtures
+
+    logger.info(f"Plotting {num_mixtures} target-truth results to {output}")
+    pdf = PdfPages(f"{output}")
+    for m in range(num_mixtures):
+        mixture_begin = mixdb["mixtures"][m]["i_sample_offset"]
+        feature_begin = mixdb["mixtures"][m]["o_frame_offset"]
+        # For each target/mixture, get index endpoint (mix,noise,target sample,...)
+        if m == num_mixtures - 1:
+            mixture_end = num_samples
+            feature_end = num_features
+        else:
+            mixture_end = mixdb["mixtures"][m + 1]["i_sample_offset"]
+            # out frames are possibly decimated/stride-combined
+            feature_end = mixdb["mixtures"][m + 1]["o_frame_offset"]
+
+        # Trim waveform data
+        mixture_trimmed = audio_data.mixture[mixture_begin:mixture_end]
+        target_trimmed = sum(audio_data.targets)[mixture_begin:mixture_end]
+
+        # Provide subset defined by mixture using truth_index array of indices (multichannel truth)
+        target_record = mixdb["targets"][mixdb["mixtures"][m]["target_id"]]
+        truth_settings = target_record["truth_settings"]
+        if len(truth_settings) > 1:
+            raise SonusAIError("Number of truth_settings is more than one. This is not supported yet.")
+
+        truth_index = np.array(truth_settings[0]["index"]) - 1
+        truth_f_trimmed = feature_data.truth_f[feature_begin:feature_end, truth_index]
+
+        if predict_data.shape[0] > 0:
+            predict_data_trimmed = predict_data[feature_begin:feature_end, :]
+            # Prediction Activity analysis to select top prediction class
+            # true if active in any frame
+            predict_activity = np.any(predict_data_trimmed >= mixdb["class_weights_threshold"], axis=0)
+            predict_activity_index = np.array([i for i, x in enumerate(predict_activity) if x]) + 1
+            logger.info(
+                f'Prediction active in classes based on threshold {mixdb["class_weights_threshold"]}:\n'
+                f'{predict_activity_index}'
+            )
+            predict_mean = np.mean(predict_data_trimmed, axis=0)
+            top_active_classes = np.argsort(predict_mean)[::-1] + 1
+            logger.info(f"Top 10 active prediction classes by mean:\n{top_active_classes[0:10]}")
+            # plot most active prediction
+            predict_data_trimmed = predict_data_trimmed[:, top_active_classes[0] - 1]
+            logger.info(f"Plotting prediction class {top_active_classes[0]}")
+
+        # Setup plot of target waveform with truth on top
+        # TODO
+        #  Make a function w/args mixture_trimmed, target_trimmed, truth_f, pr, mixdat, m, truth_index, target_path
+        # calculate number of samples per frame for plotting
+        num_plot_features = truth_f_trimmed.shape[0]
+        num_plot_samples_per_feature = len(target_trimmed) // num_plot_features
+        # number of plot samples multiple of frames
+        num_plot_samples = num_plot_samples_per_feature * num_plot_features
+        # x-axis in sec
+        x_seconds = np.arange(num_plot_samples, dtype=np.float32) / sonusai.mixture.SAMPLE_RATE
+        # Reshape/extend truth to #samples in waveform
+        num_plot_classes = truth_f_trimmed.shape[1]  # Number of plot truth classes
+        y_truth_f = np.reshape(
+            np.tile(np.expand_dims(truth_f_trimmed, 1), [1, num_plot_samples_per_feature, 1]),
+            [num_plot_samples, num_plot_classes],
+        )
+
+        fig, ax0 = plt.subplots(1, 1, constrained_layout=True, figsize=figsize)
+        ax = np.array([ax0], dtype=object)
+        plots = []
+
+        # Plot the time-domain waveforms then truth/prediction on second axis
+        if mixture_trimmed.shape[0] > 0:
+            color = "mistyrose"
+            (mix_plot,) = ax[0].plot(x_seconds, mixture_trimmed[0:num_plot_samples], color=color, label="MIX")
+            ax[0].tick_params(axis="y", labelcolor="red")
+            plots.append(mix_plot)
+
+        if target_trimmed.shape[0] > 0:
+            color = "tab:blue"
+            (tt_plot,) = ax[0].plot(x_seconds, target_trimmed[0:num_plot_samples], color=color, label="TAR")
+            ax[0].set_ylabel("ampl", color=color)
+            ax[0].tick_params(axis="y", labelcolor=color)
+            plots.append(tt_plot)
+
+        ax2 = ax[0].twinx()  # instantiate 2nd y-axis that shares the same x-axis
+
+        # Plot first truth
+        # TODO Support multi-channel
+        if truth_f_trimmed.shape[0] > 0:
+            color = "tab:green"
+            label2 = f"truth{truth_index[0] + 1}"
+            # we already handled the x-label with ax1
+            ax2.set_ylabel(label2, color=color)
+            # logger.info(f'Mixture num {m}, truth_index={truth_index}')
+            (tr_plot,) = ax2.plot(x_seconds, y_truth_f[:, 0], color=color, label=label2)
+            ax2.set_ylim([-0.05, 1.05])
+            ax2.tick_params(axis="y", labelcolor=color)
+            plots.append(tr_plot)
+
+        if predict_data.shape[0] > 0:
+            color = "tab:brown"
+            label2 = f"prcl{top_active_classes[0]}"
+            # we already handled the x-label with ax1
+            ax2.set_ylabel(label2, color=color)
+            predict_extended = np.reshape(
+                np.tile(
+                    np.expand_dims(predict_data_trimmed, 1),
+                    [1, num_plot_samples_per_feature, 1],
+                ),
+                [num_plot_samples, num_plot_classes],
+            )
+            (pr_plot,) = ax2.plot(x_seconds, predict_extended, color=color, label=label2)
+            ax2.set_ylim([-0.05, 1.05])
+            ax2.tick_params(axis="y", labelcolor=color)
+            plots.append(pr_plot)
+
+        ax[0].set_xlabel("time (s)")  # set only on last/bottom plot
+
+        # Get actual mixture target config parameters
+        target_augmentations = mixdb["target_augmentations"][mixdb["mixtures"][m]["target_augmentation_index"]]
+        fig.suptitle(
+            f'{m + 1} of {num_mixtures}: {target_path}\n'
+            f'{target_record["name"]}\n'
+            f'Target augmentations: {target_augmentations}\n'
+            f'Truth indices: {truth_settings[0]["index"]}\n'
+            f'Global Truth Function:Config {truth_settings[0]["function"]} : {truth_settings[0]["config"]}',
+            fontsize=10,
+        )
+        pdf.savefig(fig)
+        plt.close(fig)
+
+    # open config file read in as text, then print to last page
+    last_page = plt.figure(figsize=figsize)
+    last_page.clf()
+
+    option_text = (
+        f"Function parameters:\n"
+        f"  root     {root}\n"
+        f"  config   {config_file}\n"
+        f"  target   {target_file}\n"
+        f"  model    {model_file}\n"
+        f"  feature  {feature}\n"
+        f"  sedthr   {thresholds}\n\n"
+        f"----------------------------------------\n\n"
+    )
+    last_page.text(
+        0.05,
+        0.95,
+        option_text + yaml.dump(config),
+        transform=last_page.transFigure,
+        family="monospace",
+        size=10,
+        ha="left",
+        va="top",
+    )
+    pdf.savefig()
+    plt.close()
+    pdf.close()
+
+
+def main() -> None:
+    from docopt import docopt
+
+    import sonusai
+    from sonusai import create_file_handler
+    from sonusai import initial_log_messages
+    from sonusai import update_console_handler
+    from sonusai.utils.docstring import trim_docstring
+
+    args = docopt(trim_docstring(__doc__), version=sonusai.__version__, options_first=True)
+
+    verbose = args["--verbose"]
+    create_file_handler("tplot.log")
+    update_console_handler(verbose)
+    initial_log_messages("tplot")
+
+    tplot(
+        root=args["--root"],
+        config_file=args["--config"],
+        target_file=args["--target"],
+        model_file=args["--model"],
+        feature=args["--feature"],
+        thresholds=args["--sedthr"],
+        verbose=verbose,
+    )
+
+
+if __name__ == "__main__":
+    main()

sonusai/doc.py

@@ -0,0 +1,52 @@
+"""sonusai doc
+
+usage: doc [-h] [TOPIC]
+
+options:
+   -h, --help   Display this help.
+
+Show SonusAI documentation.
+
+"""
+
+
+def main() -> None:
+    from docopt import docopt
+
+    from sonusai import __version__ as sai_version
+    from sonusai.utils.docstring import trim_docstring
+
+    args = docopt(trim_docstring(__doc__), version=sai_version, options_first=True)
+
+    from sonusai import doc_strings
+
+    topic = args["TOPIC"]
+
+    print(f"SonusAI {sai_version} Documentation")
+    print("")
+
+    topics = sorted([item[4:] for item in dir(doc_strings) if item.startswith("doc_")])
+
+    if topic not in topics:
+        if topic is not None:
+            print(f"Unknown topic: {topic}")
+            print("")
+
+        print("Available topics:")
+        for item in topics:
+            print(f"  {item}")
+        return
+
+    text = getattr(doc_strings, "doc_" + topic)()
+    print(text[1:])
+
+
+if __name__ == "__main__":
+    from sonusai import exception_handler
+    from sonusai.utils.keyboard_interrupt import register_keyboard_interrupt
+
+    register_keyboard_interrupt()
+    try:
+        main()
+    except Exception as e:
+        exception_handler(e)

sonusai/doc_strings/__init__.py

@@ -0,0 +1 @@
+from .doc_strings import *  # noqa: F403