Upload grpo_training.py with huggingface_hub

grpo_training.py

@@ -31,7 +31,7 @@ if tokenizer.pad_token is None:
 # Load the model explicitly
 model = AutoModelForCausalLM.from_pretrained(
     model_name,
-    torch_dtype=torch.bfloat16 if torch.cuda.is_bf16_supported() else torch.float16,
+    dtype=torch.bfloat16 if torch.cuda.is_bf16_supported() else torch.float16,
     device_map="auto"
 )
 
@@ -87,7 +87,6 @@ def preference_reward_func(samples):
 # Initialize GRPO trainer
 trainer = GRPOTrainer(
     model=model,
-    tokenizer=tokenizer,
     reward_funcs=[preference_reward_func],
     train_dataset=train_dataset,
     eval_dataset=eval_dataset,