resolve redundancies

litgpt/__main__.py

@@ -31,16 +31,6 @@
     from jsonargparse import ArgumentParser
 
 
-def _new_parser(**kwargs: Any) -> "ArgumentParser":
-    from jsonargparse import ActionConfigFile, ArgumentParser
-
-    parser = ArgumentParser(**kwargs)
-    parser.add_argument(
-        "-c", "--config", action=ActionConfigFile, help="Path to a configuration file in json or yaml format."
-    )
-    return parser
-
-
 def main() -> None:
     parser_data = {
         "download": {"fn": download_fn, "_help": "Download weights or tokenizer data from the Hugging Face Hub."},

litgpt/finetune/adapter.py

@@ -39,7 +39,7 @@
 
 
 def setup(
-    checkpoint_dir: str,
+    checkpoint_dir: Path,
     out_dir: Path = Path("out/finetune/adapter"),
     precision: Optional[str] = None,
     quantize: Optional[Literal["bnb.nf4", "bnb.nf4-dq", "bnb.fp4", "bnb.fp4-dq", "bnb.int8-training"]] = None,
@@ -80,7 +80,6 @@ def setup(
     devices = parse_devices(devices)
     out_dir = init_out_dir(out_dir)
 
-    checkpoint_dir = Path(checkpoint_dir)
     check_valid_checkpoint_dir(checkpoint_dir)
     config = Config.from_file(checkpoint_dir / "model_config.yaml")
 

litgpt/finetune/adapter_v2.py

@@ -39,7 +39,7 @@
 
 
 def setup(
-    checkpoint_dir: str,
+    checkpoint_dir: Path,
     out_dir: Path = Path("out/finetune/adapter-v2"),
     precision: Optional[str] = None,
     quantize: Optional[Literal["bnb.nf4", "bnb.nf4-dq", "bnb.fp4", "bnb.fp4-dq", "bnb.int8-training"]] = None,
@@ -81,7 +81,6 @@ def setup(
     devices = parse_devices(devices)
     out_dir = init_out_dir(out_dir)
 
-    checkpoint_dir = Path(checkpoint_dir)
     check_valid_checkpoint_dir(checkpoint_dir)
     config = Config.from_file(checkpoint_dir / "model_config.yaml")
 

litgpt/finetune/full.py

@@ -36,7 +36,7 @@
 
 
 def setup(
-    checkpoint_dir: str,
+    checkpoint_dir: Path,
     out_dir: Path = Path("out/finetune/full"),
     precision: Optional[str] = None,
     devices: Union[int, str] = 1,
@@ -78,7 +78,6 @@ def setup(
     devices = parse_devices(devices)
     out_dir = init_out_dir(out_dir)
 
-    checkpoint_dir = Path(checkpoint_dir)
     check_valid_checkpoint_dir(checkpoint_dir)
     config = Config.from_file(checkpoint_dir / "model_config.yaml")
 

litgpt/finetune/lora.py

@@ -40,7 +40,7 @@
 
 
 def setup(
-    checkpoint_dir: str,
+    checkpoint_dir: Path,
     out_dir: Path = Path("out/finetune/lora"),
     precision: Optional[str] = None,
     quantize: Optional[Literal["bnb.nf4", "bnb.nf4-dq", "bnb.fp4", "bnb.fp4-dq", "bnb.int8-training"]] = None,
@@ -99,7 +99,6 @@ def setup(
     devices = parse_devices(devices)
     out_dir = init_out_dir(out_dir)
 
-    checkpoint_dir = Path(checkpoint_dir)
     check_valid_checkpoint_dir(checkpoint_dir)
     config = Config.from_file(
         checkpoint_dir / "model_config.yaml",

litgpt/generate/adapter.py

@@ -17,7 +17,7 @@
 
 
 def main(
-    checkpoint_dir: str,
+    checkpoint_dir: Path,
     prompt: str = "What food do llamas eat?",
     input: str = "",
     adapter_path: Path = Path("out/finetune/adapter/final/lit_model.pth.adapter"),
@@ -63,7 +63,6 @@ def main(
             samples.
         precision: Indicates the Fabric precision setting to use.
     """
-    checkpoint_dir = Path(checkpoint_dir)
     precision = precision or get_default_supported_precision(training=False)
 
     plugins = None

litgpt/generate/adapter_v2.py

@@ -17,7 +17,7 @@
 
 
 def main(
-    checkpoint_dir: str,
+    checkpoint_dir: Path,
     prompt: str = "What food do llamas eat?",
     input: str = "",
     adapter_path: Path = Path("out/finetune/adapter-v2/final/lit_model.pth.adapter_v2"),
@@ -63,7 +63,6 @@ def main(
             samples.
         precision: Indicates the Fabric precision setting to use.
     """
-    checkpoint_dir = Path(checkpoint_dir)
     precision = precision or get_default_supported_precision(training=False)
 
     plugins = None

litgpt/generate/base.py

@@ -133,7 +133,7 @@ def generate(
 
 @torch.inference_mode()
 def main(
-    checkpoint_dir: str,
+    checkpoint_dir: Path,
     prompt: str = "What food do llamas eat?",
     *,
     num_samples: int = 1,
@@ -178,7 +178,6 @@ def main(
         precision: Indicates the Fabric precision setting to use.
         compile: Whether to compile the model.
     """
-    checkpoint_dir = Path(checkpoint_dir)
     precision = precision or get_default_supported_precision(training=False)
 
     plugins = None

litgpt/generate/full.py

@@ -16,7 +16,7 @@
 
 
 def main(
-    checkpoint_dir: str,
+    checkpoint_dir: Path,
     prompt: str = "What food do llamas eat?",
     input: str = "",
     finetuned_path: Path = Path("out/full/alpaca/lit_model_finetuned.pth"),
@@ -62,7 +62,6 @@ def main(
             samples.
         precision: Indicates the Fabric precision setting to use.
     """
-    checkpoint_dir = Path(checkpoint_dir)
     precision = precision or get_default_supported_precision(training=False)
 
     plugins = None

litgpt/generate/sequentially.py

@@ -112,7 +112,7 @@ def replace_device(module: torch.nn.Module, replace: torch.device, by: torch.dev
 
 @torch.inference_mode()
 def main(
-    checkpoint_dir: str,
+    checkpoint_dir: Path,
     prompt: str = "What food do llamas eat?",
     *,
     num_samples: int = 1,
@@ -156,7 +156,6 @@ def main(
         precision: Indicates the Fabric precision setting to use.
         compile: Whether to compile the model.
     """
-    checkpoint_dir = Path(checkpoint_dir)
     precision = precision or get_default_supported_precision(training=False)
 
     plugins = None

litgpt/generate/tp.py

@@ -90,7 +90,7 @@ def tensor_parallel(fabric: L.Fabric, model: GPT) -> GPT:
 
 @torch.inference_mode()
 def main(
-    checkpoint_dir: str,
+    checkpoint_dir: Path,
     prompt: str = "What food do llamas eat?",
     *,
     num_samples: int = 1,

litgpt/scripts/convert_hf_checkpoint.py

@@ -287,7 +287,7 @@ def load_param(param: Union[torch.Tensor, NotYetLoadedTensor], name: str, dtype:
 
 @torch.inference_mode()
 def convert_hf_checkpoint(
-    checkpoint_dir: str,
+    checkpoint_dir: Path,
     *,
     model_name: Optional[str] = None,
     dtype: Optional[str] = None,
@@ -302,7 +302,6 @@ def convert_hf_checkpoint(
         dtype: The data type to convert the checkpoint files to. If not specified, the weights will remain in the
             dtype they are downloaded in.
     """
-    checkpoint_dir = Path(checkpoint_dir)
     if model_name is None:
         model_name = checkpoint_dir.name
     if dtype is not None:

litgpt/scripts/merge_lora.py

@@ -13,7 +13,7 @@
 
 
 def merge_lora(
-    checkpoint_dir: str,
+    checkpoint_dir: Path,
     pretrained_checkpoint_dir: Optional[Path] = None,
     precision: Optional[str] = None
 ) -> None:
@@ -34,7 +34,6 @@ def merge_lora(
         precision: Optional precision setting to instantiate the model weights in. By default, this will
             automatically be inferred from the metadata in the given ``checkpoint_dir`` directory.
     """
-    checkpoint_dir = Path(checkpoint_dir)
     check_valid_checkpoint_dir(checkpoint_dir, model_filename="lit_model.pth.lora")
     if pretrained_checkpoint_dir is not None:
         check_valid_checkpoint_dir(pretrained_checkpoint_dir)