switching to high quality piper tts and added label translations

venv_piper/lib/python3.12/site-packages/piper/train/__init__.py

@@ -0,0 +1 @@
+"""Training code for Piper 1."""

venv_piper/lib/python3.12/site-packages/piper/train/__main__.py

@@ -0,0 +1,38 @@
+import logging
+
+import torch
+from lightning.pytorch.cli import LightningCLI
+
+from .vits.dataset import VitsDataModule
+from .vits.lightning import VitsModel
+
+_LOGGER = logging.getLogger(__package__)
+
+
+class VitsLightningCLI(LightningCLI):
+    def add_arguments_to_parser(self, parser):
+        parser.link_arguments("data.batch_size", "model.batch_size")
+        parser.link_arguments("data.num_symbols", "model.num_symbols")
+        parser.link_arguments("model.num_speakers", "data.num_speakers")
+        parser.link_arguments("model.sample_rate", "data.sample_rate")
+        parser.link_arguments("model.filter_length", "data.filter_length")
+        parser.link_arguments("model.hop_length", "data.hop_length")
+        parser.link_arguments("model.win_length", "data.win_length")
+        parser.link_arguments("model.segment_size", "data.segment_size")
+
+
+def main():
+    logging.basicConfig(level=logging.INFO)
+    torch.backends.cuda.matmul.allow_tf32 = True
+    torch.backends.cudnn.allow_tf32 = True
+    torch.backends.cudnn.deterministic = False
+    _cli = VitsLightningCLI(  # noqa: ignore=F841
+        VitsModel, VitsDataModule, trainer_defaults={"max_epochs": -1}
+    )
+
+
+# -----------------------------------------------------------------------------
+
+
+if __name__ == "__main__":
+    main()

venv_piper/lib/python3.12/site-packages/piper/train/export_generator.py

@@ -0,0 +1,62 @@
+#!/usr/bin/env python3
+
+import argparse
+import logging
+from pathlib import Path
+
+import torch
+
+from .vits.lightning import VitsModel
+
+_LOGGER = logging.getLogger(__name__)
+
+
+def main() -> None:
+    """Main entry point"""
+    torch.manual_seed(1234)
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--checkpoint", required=True, help="Path to model checkpoint (.ckpt)"
+    )
+    parser.add_argument("--generator", required=True, help="Path to output file (.pt)")
+
+    parser.add_argument(
+        "--debug", action="store_true", help="Print DEBUG messages to the console"
+    )
+    args = parser.parse_args()
+
+    if args.debug:
+        logging.basicConfig(level=logging.DEBUG)
+    else:
+        logging.basicConfig(level=logging.INFO)
+
+    _LOGGER.debug(args)
+
+    # -------------------------------------------------------------------------
+
+    generator_path = Path(args.generator)
+    generator_path.parent.mkdir(parents=True, exist_ok=True)
+
+    checkpoint_path = Path(args.checkpoint)
+
+    # pylint: disable=no-value-for-parameter
+    model = VitsModel.load_from_checkpoint(checkpoint_path, map_location="cpu")
+    model_g = model.model_g
+
+    # Inference only
+    model_g.eval()
+
+    with torch.no_grad():
+        model_g.dec.remove_weight_norm()
+
+    model_g.forward = model_g.infer  # type: ignore[method-assign,assignment]
+
+    torch.save(model_g, generator_path)
+    _LOGGER.info("Exported generator to %s", generator_path)
+
+
+# -----------------------------------------------------------------------------
+
+if __name__ == "__main__":
+    main()

venv_piper/lib/python3.12/site-packages/piper/train/export_onnx.py

@@ -0,0 +1,112 @@
+#!/usr/bin/env python3
+
+import argparse
+import logging
+from pathlib import Path
+from typing import Optional
+
+import torch
+
+from .vits.lightning import VitsModel
+
+_LOGGER = logging.getLogger(__name__)
+OPSET_VERSION = 15
+
+
+def main() -> None:
+    """Main entry point"""
+    torch.manual_seed(1234)
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--checkpoint", required=True, help="Path to model checkpoint (.ckpt)"
+    )
+    parser.add_argument(
+        "--output-file", required=True, help="Path to output file (.onnx)"
+    )
+
+    parser.add_argument(
+        "--debug", action="store_true", help="Print DEBUG messages to the console"
+    )
+    args = parser.parse_args()
+
+    if args.debug:
+        logging.basicConfig(level=logging.DEBUG)
+    else:
+        logging.basicConfig(level=logging.INFO)
+
+    _LOGGER.debug(args)
+
+    # -------------------------------------------------------------------------
+
+    output_path = Path(args.output_file)
+    output_path.parent.mkdir(parents=True, exist_ok=True)
+
+    checkpoint_path = Path(args.checkpoint)
+
+    # pylint: disable=no-value-for-parameter
+    model = VitsModel.load_from_checkpoint(checkpoint_path, map_location="cpu")
+    model_g = model.model_g
+
+    # Inference only
+    model_g.eval()
+
+    with torch.no_grad():
+        model_g.dec.remove_weight_norm()
+
+    def infer_forward(text, text_lengths, scales, sid=None):
+        noise_scale = scales[0]
+        length_scale = scales[1]
+        noise_scale_w = scales[2]
+        audio = model_g.infer(
+            text,
+            text_lengths,
+            noise_scale=noise_scale,
+            length_scale=length_scale,
+            noise_scale_w=noise_scale_w,
+            sid=sid,
+        )[0].unsqueeze(1)
+
+        return audio
+
+    model_g.forward = infer_forward  # type: ignore[method-assign,assignment]
+
+    num_symbols = model_g.n_vocab
+    num_speakers = model_g.n_speakers
+
+    dummy_input_length = 50
+    sequences = torch.randint(
+        low=0, high=num_symbols, size=(1, dummy_input_length), dtype=torch.long
+    )
+    sequence_lengths = torch.LongTensor([sequences.size(1)])
+
+    sid: Optional[torch.LongTensor] = None
+    if num_speakers > 1:
+        sid = torch.LongTensor([0])
+
+    # noise, noise_w, length
+    scales = torch.FloatTensor([0.667, 1.0, 0.8])
+    dummy_input = (sequences, sequence_lengths, scales, sid)
+
+    # Export
+    torch.onnx.export(
+        model=model_g,
+        args=dummy_input,
+        f=output_path,
+        verbose=False,
+        opset_version=OPSET_VERSION,
+        input_names=["input", "input_lengths", "scales", "sid"],
+        output_names=["output"],
+        dynamic_axes={
+            "input": {0: "batch_size", 1: "phonemes"},
+            "input_lengths": {0: "batch_size"},
+            "output": {0: "batch_size", 2: "time"},
+        },
+    )
+    _LOGGER.info("Exported model to %s", output_path)
+
+
+# -----------------------------------------------------------------------------
+
+if __name__ == "__main__":
+    main()