fix: improve code quality and fix bugs

- Fix critical typo: torch.optim.adam -> torch.optim.Adam
- Improve exception handling: bare except -> specific json.JSONDecodeError
- Add proper NotImplementedError messages for placeholder classes
- Add missing DatasetDict type hint in text_dataset.py
- Add reference to AGENTS.md in CONTRIBUTING.md

Author: glennko

CONTRIBUTING.md

@@ -2,6 +2,8 @@
 
 We welcome and appreciate contributions to xTuring! Whether it's a bug fix, a new feature, or simply a typo, every little bit helps.
 
+Before starting, please skim the [Repository Guidelines](AGENTS.md) for project structure, local commands, style, and testing conventions.
+
 ## Getting Started
 
 1. Fork the repository on GitHub

src/xturing/datasets/text2image_dataset.py

@@ -6,7 +6,7 @@ class Text2ImageDataset:
     config_name: str = "text2image_dataset"
 
     def __init__(self, path: Union[str, Path]):
-        pass
+        raise NotImplementedError("Text2ImageDataset is not implemented yet.")
 
     def _validate(self):
-        pass
+        raise NotImplementedError

src/xturing/datasets/text_dataset.py

@@ -16,7 +16,7 @@ class TextDatasetMeta:
 class TextDataset(BaseDataset):
     config_name: str = "text_dataset"
 
-    def __init__(self, path: Union[str, Path, HFDataset, dict]):
+    def __init__(self, path: Union[str, Path, HFDataset, DatasetDict, dict]):
         if isinstance(path, HFDataset) or isinstance(path, DatasetDict):
             self.data = path
         elif isinstance(path, dict):

src/xturing/models/stable_diffusion.py

@@ -8,22 +8,25 @@ class StableDiffusion:
     config_name: str = "stable_diffusion"
 
     def __init__(self, weights_path: str):
-        pass
+        raise NotImplementedError(
+            "StableDiffusion is a placeholder and not yet implemented."
+        )
 
     def finetune(self, dataset: Text2ImageDataset, logger=True):
         """Finetune Stable Diffusion model on a given dataset.
-        
+
         Args:
             dataset (Text2ImageDataset): Dataset to finetune on.
-            logger (bool, optional): To be setup with a Pytorch Lightning logger when implemented."""
-        pass
+            logger (bool, optional): To be setup with a Pytorch Lightning logger when implemented.
+        """
+        raise NotImplementedError
 
     def generate(
         self,
         texts: Optional[Union[List[str], str]] = None,
         dataset: Optional[Text2ImageDataset] = None,
     ):
-        pass
+        raise NotImplementedError
 
     def save(self, path: Union[str, Path]):
-        pass
+        raise NotImplementedError

src/xturing/self_instruct/generate_instances.py

@@ -62,8 +62,9 @@ def generate_instances(
                 try:
                     data = json.loads(line)
                     existing_requests[data["instruction"]] = data
-                except:
-                    pass
+                except json.JSONDecodeError:
+                    # Skip malformed JSON lines
+                    continue
         print(f"Loaded {len(existing_requests)} existing requests")
 
     progress_bar = tqdm(total=len(tasks))

src/xturing/self_instruct/identify_if_classification.py

@@ -40,8 +40,9 @@ def identify_if_classification(
                 try:
                     data = json.loads(line)
                     existing_requests[data["instruction"]] = data
-                except:
-                    pass
+                except json.JSONDecodeError:
+                    # Skip malformed JSON lines
+                    continue
         print(f"Loaded {len(existing_requests)} existing requests")
 
     # Create the progress bar

src/xturing/trainers/lightning_trainer.py

@@ -47,7 +47,7 @@ def configure_optimizers(self):
                 self.pytorch_model.parameters(), lr=self.learning_rate
             )
         elif self.optimizer_name == "adam":
-            optimizer = torch.optim.adam(
+            optimizer = torch.optim.Adam(
                 self.pytorch_model.parameters(), lr=self.learning_rate
             )
         elif self.optimizer_name == "cpu_adam":