feat: handling wrong model type

hfgl/cli.py

@@ -130,15 +130,22 @@ def synthesize(
     ),
 ):
     """Given some Mel spectrograms and a trained model, generate some audio. i.e. perform *copy synthesis*"""
+    import sys
+
     import torch
+    from pydantic import ValidationError
     from scipy.io.wavfile import write
 
     from .utils import load_hifigan_from_checkpoint, synthesize_data
 
     device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
     checkpoint = torch.load(generator_path, map_location=device)
     data = torch.load(data_path, map_location=device)
-    vocoder_model, vocoder_config = load_hifigan_from_checkpoint(checkpoint, device)
+    try:
+        vocoder_model, vocoder_config = load_hifigan_from_checkpoint(checkpoint, device)
+    except (TypeError, ValidationError) as e:
+        logger.error(f"Unable to load {generator_path}: {e}")
+        sys.exit(1)
     wav, sr = synthesize_data(data, vocoder_model, vocoder_config)
     logger.info(f"Writing file {data_path}.wav")
     write(f"{data_path}.wav", sr, wav)

hfgl/model.py

@@ -11,6 +11,8 @@
     dynamic_range_compression_torch,
     get_spectral_transform,
 )
+from loguru import logger
+from pydantic import ValidationError
 from torch.nn import AvgPool1d, Conv1d, Conv2d, ConvTranspose1d
 from torch.nn.utils import spectral_norm
 from torch.nn.utils.parametrizations import weight_norm
@@ -454,8 +456,16 @@ class HiFiGANGenerator(pl.LightningModule):
 
     def __init__(self, config: dict | VocoderConfig):
         super().__init__()
+
         if not isinstance(config, VocoderConfig):
-            config = VocoderConfig(**config)
+            try:
+                config = VocoderConfig(**config)
+            except ValidationError as e:
+                logger.error(f"{e}")
+                raise TypeError(
+                    "Unable to load config.  Possible causes: is it really a VocoderConfig? or the correct version?"
+                )
+
         self.config = config
         self.generator = Generator(self.config)
 
@@ -464,7 +474,14 @@ def on_load_checkpoint(self, checkpoint):
         Note, this shouldn't fail on different versions of pydantic anymore,
         but it will fail on breaking changes to the config. We should catch those exceptions
         and handle them appropriately."""
-        self.config = VocoderConfig(**checkpoint["hyper_parameters"]["config"])
+        try:
+            config = VocoderConfig(**checkpoint["hyper_parameters"]["config"])
+        except ValidationError as e:
+            logger.error(f"{e}")
+            raise TypeError(
+                "Unable to load config.  Possible causes: is it really a VocoderConfig? or the correct version?"
+            )
+        self.config = config
 
     def on_save_checkpoint(self, checkpoint):
         """Serialize the checkpoint hyperparameters"""
@@ -479,7 +496,13 @@ def __init__(self, config: dict | VocoderConfig):
         # Because we serialize the configurations when saving checkpoints,
         # sometimes what is passed is actually just a dict.
         if not isinstance(config, VocoderConfig):
-            config = VocoderConfig(**config)
+            try:
+                config = VocoderConfig(**config)
+            except ValidationError as e:
+                logger.error(f"{e}")
+                raise TypeError(
+                    "Unable to load config.  Possible causes: is it really a VocoderConfig? or the correct version?"
+                )
         self.config = config
         self.mpd = MultiPeriodDiscriminator(config)
         self.msd = MultiScaleDiscriminator(config)

hfgl/utils.py

@@ -4,6 +4,7 @@
 
 import numpy as np
 import torch
+from everyvoice.utils import pydantic_validation_error_shortener
 from everyvoice.utils.heavy import get_spectral_transform
 from loguru import logger
 
@@ -23,7 +24,15 @@ def sizeof_fmt(num, suffix="B"):
 def load_hifigan_from_checkpoint(ckpt: dict, device) -> Tuple[HiFiGAN, HiFiGANConfig]:
     config: dict | HiFiGANConfig = ckpt["hyper_parameters"]["config"]
     if isinstance(config, dict):
-        config = HiFiGANConfig(**config)
+        from pydantic import ValidationError
+
+        try:
+            config = HiFiGANConfig(**config)
+        except ValidationError as e:
+            logger.error(f"{pydantic_validation_error_shortener(e)}")
+            raise TypeError(
+                "Unable to load config.  Possible causes: is it really a VocoderConfig? or the correct version?"
+            )
     if any(
         (
             key.startswith("mpd") or key.startswith("msd")