refactor: switch steps back to option

remove named configurations

everyvoice/base_cli/helpers.py

@@ -35,39 +35,26 @@
 
 
 def load_config_base_command(
-    name: Enum,
     model_config: Union[
         DFAlignerConfig, EveryVoiceConfig, FastSpeech2Config, HiFiGANConfig
     ],
-    configs,
     # Must include the above in model-specific command
     config_args: List[str],
     config_path: Path,
 ):
     from everyvoice.utils import update_config_from_cli_args
 
-    if config_path:
-        config = model_config.load_config_from_path(config_path)
-    elif name:
-        config = model_config.load_config_from_path(configs[name.value])
-    else:
-        logger.error(
-            "You must either choose a <NAME> of a preconfigured dataset, or provide a <CONFIG_PATH> to a preprocessing configuration file."
-        )
-        exit()
+    config = model_config.load_config_from_path(config_path)
 
     config = update_config_from_cli_args(config_args, config)
     return config
 
 
 def preprocess_base_command(
-    name: Enum,
     model_config: Union[
         DFAlignerConfig, EveryVoiceConfig, FastSpeech2Config, HiFiGANConfig
     ],
-    configs,
-    steps,
-    preprocess_categories,
+    steps: List[str],
     # Must include the above in model-specific command
     config_args: List[str],
     config_path: Path,
@@ -78,34 +65,24 @@ def preprocess_base_command(
 ):
     from everyvoice.preprocessor import Preprocessor
 
-    config = load_config_base_command(
-        name, model_config, configs, config_args, config_path
-    )
-    to_process = [x.name for x in steps]
+    config = load_config_base_command(model_config, config_args, config_path)
     preprocessor = Preprocessor(config)
-    if not steps:
-        logger.info(
-            f"No specific preprocessing data requested, processing everything from dataset '{name}'"
-        )
-        to_process = list(preprocess_categories.__members__.keys())
     if isinstance(config, FastSpeech2Config) and config.model.use_phonological_feats:
-        to_process.append("pfs")
+        steps.append("pfs")
     preprocessor.preprocess(
         output_path=output_path,
         cpus=cpus,
         overwrite=overwrite,
-        to_process=to_process,
+        to_process=steps,
         debug=debug,
     )
-    return preprocessor, config, to_process
+    return preprocessor, config, steps
 
 
 def train_base_command(
-    name: Enum,
     model_config: Union[
         DFAlignerConfig, EveryVoiceConfig, FastSpeech2Config, HiFiGANConfig
     ],
-    configs,
     data_module: Union[
         AlignerDataModule, E2EDataModule, FastSpeech2DataModule, HiFiGANDataModule
     ],
@@ -119,9 +96,7 @@ def train_base_command(
     nodes: int,
     strategy: str,
 ):
-    config = load_config_base_command(
-        name, model_config, configs, config_args, config_path
-    )
+    config = load_config_base_command(model_config, config_args, config_path)
     logger.info("Loading modules for training...")
     pbar = tqdm(range(4))
     pbar.set_description("Loading pytorch and friends")

everyvoice/base_cli/interfaces.py

@@ -21,22 +21,44 @@ def load_config_base_command_interface(
 
 
 def preprocess_base_command_interface(
-    config_args: List[str] = typer.Option(None, "--config-args", "-c"),
-    config_path: Path = typer.Option(
-        None, "--config-path", "-p", exists=True, dir_okay=False, file_okay=True
+    config_path: Path = typer.Argument(
+        ...,
+        exists=True,
+        dir_okay=False,
+        file_okay=True,
+        help="The path to your model configuration file.",
+    ),
+    config_args: List[str] = typer.Option(
+        None, "-c", "--config-args", help="Overwrite the configuration"
+    ),
+    output_path: Optional[Path] = typer.Option(
+        "filelist.psv",
+        "-o",
+        "--output",
+        help="The path to where the processed data filelist should be written",
+    ),
+    cpus: Optional[int] = typer.Option(
+        min(4, mp.cpu_count()),
+        "-C",
+        "--cpus",
+        help="How many CPUs to use when preprocessing",
     ),
-    output_path: Optional[Path] = typer.Option("filelist.psv", "-o", "--output"),
-    cpus: Optional[int] = typer.Option(min(4, mp.cpu_count()), "-C", "--cpus"),
     overwrite: bool = typer.Option(False, "-O", "--overwrite"),
     debug: bool = typer.Option(False, "-D", "--debug"),
 ):
     pass
 
 
 def train_base_command_interface(
-    config_args: List[str] = typer.Option(None, "--config", "-c"),
-    config_path: Path = typer.Option(
-        None, "--config-path", "-p", exists=True, dir_okay=False, file_okay=True
+    config_path: Path = typer.Argument(
+        ...,
+        exists=True,
+        dir_okay=False,
+        file_okay=True,
+        help="The path to your model configuration file.",
+    ),
+    config_args: List[str] = typer.Option(
+        None, "-c", "--config-args", help="Overwrite the configuration"
     ),
     accelerator: str = typer.Option(
         "auto",

everyvoice/cli.py

@@ -2,7 +2,6 @@
 
 import typer
 
-from everyvoice.config import CONFIGS
 from everyvoice.model.feature_prediction.FastSpeech2_lightning.fs2.cli import (
     preprocess as preprocess_fs2,
 )
@@ -101,7 +100,7 @@ def new_dataset():
 
     This command will preprocess all of the data you need for use with EveryVoice.
 
-    By default every step of the preprocessor will be done, but you can run specific commands by adding them as arguments for example: **everyvoice preprocess energy pitch**
+    By default every step of the preprocessor will be done, but you can run specific commands by adding them as options for example: **everyvoice preprocess path/to/config.yaml -s energy -s pitch**
     """,
 )(preprocess_fs2)
 
@@ -145,22 +144,18 @@ def new_dataset():
     help="""
     # Synthesize Help
 
-        - **text-to-spec** --- this is the most common model to run
+        - **text-to-spec** --- this is the most common model to run for performing normal speech synthesis.
 
-        - **spec-to-wav** --- this is the model that turns your spectral features into audio. It is also known as a 'vocoder'. You will typically not need to train your own version. Please refer to [https://pathtocheckpoints](https://pathtocheckpoints) for more information.
+        - **spec-to-wav** --- this is the model that turns your spectral features into audio. this type of synthesis is also known as copy synthesis and unless you know what you are doing, you probably don't want to do this.
     """,
 )
 
 synthesize_group.command(
     name="text-to-wav",
-    short_help="Given some text and a trained model, generate some audio",
-    help="Given some text and a trained model, generate some audio.",
 )(synthesize_fs2)
 
 synthesize_group.command(
     name="spec-to-wav",
-    short_help="Given some Mel spectrograms and a trained model, generate some audio",
-    help="Given some Mel spectrograms and a trained model, generate some audio.",
 )(synthesize_hfg)
 
 app.add_typer(
@@ -170,11 +165,6 @@ def new_dataset():
 )
 
 
-_config_keys = {k: k for k in CONFIGS.keys()}
-
-CONFIGS_ENUM = Enum("CONFIGS", _config_keys)  # type: ignore
-
-
 class TestSuites(str, Enum):
     all = "all"
     configs = "configs"

everyvoice/config/__init__.py

@@ -1,16 +0,0 @@
-from pathlib import Path
-from typing import Dict
-
-from everyvoice.config import __file__ as everyvoice_file
-
-CONFIGS: Dict[str, Path] = {
-    "base": Path(everyvoice_file).parent / "base" / "base_composed.yaml",
-    "lj": Path(everyvoice_file).parent / "lj" / "lj.yaml",
-    "istft": Path(everyvoice_file).parent / "lj" / "lj_istft.yaml",
-    "openslr": Path(everyvoice_file).parent / "openslr" / "openslr.yaml",
-}
-
-
-class ConfigError(Exception):
-    def __init__(self, msg):
-        super().__init__(msg)