0731_2

arealite/api/cli_args.py

@@ -637,6 +637,9 @@ class DatasetConfig:
         default=0, metadata={"help": "Number of worker processes for data loading"}
     )
     drop_last: bool = field(default=True)
+    reward_fn: Optional[str] = field(
+        default=None,
+    )
 
 
 @dataclass

examples/arealite/configs/geometry3k_grpo.yaml

@@ -99,6 +99,7 @@ train_dataset:
   num_workers: 4
   path: hiyouga/geometry3k
   type: rl
+  reward_fn: ${train_dataset.path}
 
 valid_dataset: 
   batch_size: 32
@@ -108,8 +109,8 @@ valid_dataset:
   path: ${train_dataset.path}
   type: rl
 
-reward_fn:
-  path: ${train_dataset.path}
+
+  
 
 # Utilities
 saver:
@@ -140,6 +141,5 @@ stats_logger:
   experiment_name: ${experiment_name}
   trial_name: ${trial_name}
   fileroot: ${cluster.fileroot}
-
-wandb:
-  project: geometry3k-grpo
+  wandb:
+    project: geometry3k-grpo

examples/arealite/reward/geometry3k.py

@@ -13,19 +13,20 @@ def geometry3k_reward_fn(
 ):
     sol = extract_answer(completions, data_name="")  # str number
     ans = answer
-
+    sol = sol.replace(" ", "")
+    ans= ans.replace(" ", "")
     if sol is None:
         return 0
     if ans is None:
         return 0
 
-    is_numeric = sol.replace('.', '', 1).isdigit()  # Allows for decimal check
-    is_latex = sol.startswith("\\frac") or '\\sqrt' in sol
-    print(f"completions: {completions}, answer: {answer}")
+    is_numeric = sol.replace('.', '', 1).isdigit() or ans.replace('.', '', 1).isdigit() # Allows for decimal check
+    is_latex = sol.startswith("\\frac") or '\\sqrt' in sol or ans.startswith("\\frac") or '\\sqrt' in ans
+    print(f"sol: {sol}, ans: {ans}")
     # Exact answer matching
     if sol == ans :
         reward = 1
-    elif is_numeric and abs(float(sol) - float(ans)) < 1e-4:
+    elif is_numeric and not is_latex and abs(float(sol) - float(ans)) < 1e-4:
         reward = 0.8  # Reward for correct numerical approximation
     elif is_latex:
         # Check if numbers in LaTeX are correct

examples/arealite/vision_grpo.py

@@ -28,11 +28,13 @@ from realhf.base import stats_tracker
 
 def main(args):
 
-    wandb.init(project=config.wandb.project)
+    
 
     config, _ = load_expr_config(args, GRPOConfig)
     config: GRPOConfig
 
+    wandb.init(project=config.stats_logger.wandb.project)
+
     rank = int(os.getenv("RANK"))
     world_size = int(os.getenv("WORLD_SIZE"))
     processor, tokenizer = load_hf_processor_and_tokenizer(config.tokenizer_path)
@@ -114,7 +116,7 @@ def main(args):
         config.gconfig.stop_token_ids.append(tokenizer.eos_token_id)
         
     reward_fn = custom_reward_fn(
-        path=config.reward_fn.path,
+        path=config.train_dataset.reward_fn,
     )
 
     workflow = VisionRLVRWorkflow(