codeBERT training on codeparrot-clean

code/pyproject.toml

@@ -10,15 +10,15 @@ dependencies = [
     "torch==2.5.1",
     "tqdm==4.66.5",
     "tree-sitter==0.23.1",
-    "transformers[torch]>=4.46.3",
+    "transformers[torch]==4.46.3",
     "datasets==3.0.1",
     "huggingface-hub==0.26.0",
     "matplotlib==3.9.2",
     "scikit-learn==1.5.2",
     "seaborn==0.13.2",
     "tree-sitter-python==0.23.4",
-    "ipykernel>=6.29.5",
+    "ipykernel==6.29.5",
-    "ipywidgets>=8.1.5",
+    "ipywidgets==8.1.5",
 ]
 requires-python = "==3.11.*"
 readme = "README.md"

code/src/config.json

@@ -1,17 +1,20 @@
 {
-    "extra_embeddings": true,
+    "extra_embeddings": false,
-    "run_name": "tree",
+    "run_name": "original",
-    "data_dir": "./data/CodeSearchNet-parsed/python/",
+    "data_dir": "./data/codeparrot-clean-parsed/",
-    "output_dir": "./outputs/tree",
+    "output_dir": "./outputs/original",
     "seed": 420,
     "mlm_probability": 0.15,
     "batch_size": 32,
-    "epochs": 5,
+    "epochs": 1,
-    "eval_every": 1000,
+    "eval_every": 5000,
     "learning_rate": 5e-4,
     "weight_decay": 0.1,
     "max_grad_norm": 1.0,
     "warmup_steps": 1000,
     "fp16": true,
-    "logging_steps": 100
+    "logging_steps": 100,
+    "valid_size": 0.05,
+    "test_size": 0.05,
+    "num_samples": -1
 }

code/src/parse_dataset.py

@@ -10,7 +10,6 @@ from transformers import AutoTokenizer
 import warnings
 warnings.filterwarnings("ignore", category=SyntaxWarning)
 
-# Configure logging
 logging.basicConfig(
     level=logging.INFO, 
     format='%(asctime)s - %(levelname)s - %(message)s',
@@ -78,9 +77,9 @@ def analyze_code_with_codebert_and_treesitter(code_snippet: str, tokenizer: Auto
     
     return encoded['input_ids'], encoded['attention_mask'], tokens, depths, sibling_idxs, node_texts
 
-def process_batch(examples, tokenizer) -> Dict[str, List[Any]]:
+def process_batch(examples, tokenizer, code_column) -> Dict[str, List[Any]]:
     """Process a batch of examples."""
-    contents = examples['code']
+    contents = examples[code_column]
 
     processed_input_ids = []
     processed_attention_mask = []
@@ -119,9 +118,11 @@ def process_batch(examples, tokenizer) -> Dict[str, List[Any]]:
     }
 
 def main():
+    code_column = 'content'
+
     current_dir = Path(__file__).parent
-    input_dir = current_dir.parent / 'data' / 'CodeSearchNet' / 'python'
+    input_dir = current_dir.parent / 'data' / 'codeparrot-clean'
-    output_dir = current_dir.parent / 'data' / 'CodeSearchNet-parsed' / 'python'
+    output_dir = current_dir.parent / 'data' / 'codeparrot-clean-parsed'
     output_dir.mkdir(parents=True, exist_ok=True)
     
     # Initialize tokenizer and model from scratch
@@ -132,17 +133,13 @@ def main():
     num_proc = min(multiprocessing.cpu_count() - 1, 32)
     logger.info(f"Using {num_proc} processes for dataset processing") 
     
-    for dataset_name in ['valid', 'test', 'train']:
+    dataset = load_dataset("codeparrot/codeparrot-clean", cache_dir=str(input_dir), split='train')
-        logger.info(f"Processing dataset: {dataset_name}")
-        
-        input_file = input_dir / f'{dataset_name}.jsonl'
-        dataset = load_dataset('json', data_files=str(input_file))['train']
     original_dataset_size = len(dataset)
-        logger.info(f"Loaded dataset from {input_file} with {original_dataset_size} examples")
+    logger.info(f"Loaded dataset from {input_dir} with {original_dataset_size} examples")
     
     def tokenize_function(examples):
         return tokenizer(
-                examples['code'],
+            examples[code_column],
             padding='max_length',
             truncation=True,
             max_length=512,
@@ -159,7 +156,7 @@ def main():
     
     processed_dataset = tokenized_dataset.map(
         process_batch,
-            fn_kwargs={'tokenizer': tokenizer},
+        fn_kwargs={'tokenizer': tokenizer, 'code_column': code_column},
         batched=True,
         desc="Processing",
         num_proc=num_proc,
@@ -177,11 +174,10 @@ def main():
         logger.error("No valid examples found in dataset!")
         return
     
-        output_file = output_dir / f'{dataset_name}'
+    processed_dataset.save_to_disk(str(output_dir))
-        processed_dataset.save_to_disk(str(output_file))
+    logger.info(f"Saved processed dataset to {output_dir} with {len(processed_dataset)} examples")
-        logger.info(f"Saved processed dataset to {output_file} with {len(processed_dataset)} examples")
         
-    logger.info("All datasets processed")
+    logger.info("Dataset processed")
 
 if __name__ == "__main__":
     main()

code/src/training.py

@@ -14,7 +14,11 @@ from transformers import (
 
 from tree_model import TreeCodeBERTForPreTraining
 
-logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
+logging.basicConfig(
+    level=logging.INFO, 
+    format='%(asctime)s - %(levelname)s - %(message)s',
+    datefmt='%Y-%m-%d %H:%M:%S'
+)
 logger = logging.getLogger(__name__)
 
 def load_config(config_path: Path) -> dict:
@@ -29,7 +33,7 @@ def main():
     output_dir = Path(config['output_dir'])
 
     # Initialize W&B
-    wandb.init(project='easy_training', config=config, name=config['run_name'])
+    wandb.init(project='codeparrot', config=config, name=config['run_name'])
     
     # Upload the training files to W&B
     wandb.save(__file__)
@@ -37,11 +41,25 @@ def main():
     if config['extra_embeddings']:
         wandb.save(current_dir / 'tree_model.py')
     
+    if 'CodeSearchNet' in config['data_dir']:
         dataset = DatasetDict({
             'train': load_from_disk(data_dir / 'train'),
             'valid': load_from_disk(data_dir / 'valid'),
             'test': load_from_disk(data_dir / 'test')
         })
+    else:
+        
+        dataset = load_from_disk(data_dir)
+        if config['num_samples'] > 0:
+            dataset = dataset.select(range(config['num_samples']))
+        train_testvalid = dataset.train_test_split(test_size=config['test_size'] + config['valid_size'])
+        test_valid = train_testvalid['test'].train_test_split(test_size=config['valid_size'] / (config['test_size'] + config['valid_size']))
+        dataset = DatasetDict({
+            'train': train_testvalid['train'],
+            'test': test_valid['test'],
+            'valid': test_valid['train'],
+        })
+        
     
     # Continue with the rest of processing
     columns_to_remove = dataset['train'].column_names
@@ -107,8 +125,9 @@ def main():
 
     # Evaluate
     logger.info('Evaluating on test set...')
-    eval_results = trainer.evaluate(eval_dataset=dataset['test'])
+    eval_results = trainer.evaluate(dataset['test'])
-    logger.info(eval_results)
+    wandb.log({'test_loss': eval_results['eval_loss']})
+    logger.info(f'Test loss: {eval_results["eval_loss"]}')
     
     # Save final model
     logger.info('Saving final model...')