update lora.py

llms/mlx_lm/lora.py

@@ -43,6 +43,7 @@ yaml_loader.add_implicit_resolver(
 CONFIG_DEFAULTS = {
     "model": "mlx_model",
     "train": False,
+    "training_mode": "normal",
     "fine_tune_type": "lora",
     "data": "data/",
     "seed": 0,
@@ -62,6 +63,10 @@ CONFIG_DEFAULTS = {
     "config": None,
     "grad_checkpoint": False,
     "lr_schedule": None,
+    "reference_model_path": None,
+    "group_size": 4,
+    "beta": 0.1,
+    "epsilon": 1e-4,
     "lora_parameters": {"rank": 8, "alpha": 16, "dropout": 0.0, "scale": 10.0},
 }
 
@@ -95,6 +100,12 @@ def build_parser():
         choices=["lora", "dora", "full"],
         help="Type of fine-tuning to perform: lora, dora, or full.",
     )
+    parser.add_argument(
+        "--training-mode",
+        type=str,
+        choices=["normal", "grpo"],
+        help="Training mode: normal or GRPO",
+    )
     parser.add_argument(
         "--num-layers",
         type=int,
@@ -162,6 +173,25 @@ def build_parser():
         default=None,
     )
     parser.add_argument("--seed", type=int, help="The PRNG seed")
+
+    parser.add_argument(
+        "--group-size",
+        type=int,
+        help="Number of responses per prompt.",
+        default=4,
+    )
+    parser.add_argument(
+        "--beta",
+        type=float,
+        help="KL penalty coefficient.",
+        default=0.1,
+    )
+    parser.add_argument(
+        "--epsilon",
+        type=float,
+        help="The Epsilon for numerical stability.",
+        default=1e-4,
+    )
     return parser
 
 
@@ -221,32 +251,98 @@ def train_model(
         )
     )
     # Train model
-    train(
+    if args.training_mode == "grpo":
-        model=model,
+        training_args = GRPOTrainingArgs(
-        tokenizer=tokenizer,
+            batch_size=args.batch_size,
-        args=training_args,
+            iters=args.iters,
-        optimizer=opt,
+            val_batches=args.val_batches,
-        train_dataset=train_set,
+            steps_per_report=args.steps_per_report,
-        val_dataset=valid_set,
+            steps_per_eval=args.steps_per_eval,
-        training_callback=training_callback,
+            steps_per_save=args.save_every,
-    )
+            adapter_file=adapter_file,
+            max_seq_length=args.max_seq_length,
+            grad_checkpoint=args.grad_checkpoint,
+            beta=args.beta,
+            group_size=args.group_size,
+            epsilon=args.epsilon,
+            reference_model_path=args.reference_model_path
+        )
+
+        if args.reference_model_path:
+            reference_model, _ = load(args.reference_model_path)
+        else:
+            reference_model, _ = load(args.model)
+        
+        train_grpo(
+            model=model,
+            reference_model=reference_model.freeze(),
+            tokenizer=tokenizer,
+            optimizer=opt,
+            train_dataset=train_set,
+            val_dataset=valid_set,
+            args=training_args,
+            training_callback=training_callback,
+        )
+    else:
+        training_args = TrainingArgs(
+            batch_size=args.batch_size,
+            iters=args.iters,
+            val_batches=args.val_batches,
+            steps_per_report=args.steps_per_report,
+            steps_per_eval=args.steps_per_eval,
+            steps_per_save=args.save_every,
+            adapter_file=adapter_file,
+            max_seq_length=args.max_seq_length,
+            grad_checkpoint=args.grad_checkpoint
+        )
+
+        train(
+            model=model,
+            tokenizer=tokenizer,
+            args=training_args,
+            optimizer=opt,
+            train_dataset=train_set,
+            val_dataset=valid_set,
+            training_callback=training_callback,
+        )
 
 
 def evaluate_model(args, model: nn.Module, tokenizer: TokenizerWrapper, test_set):
     model.eval()
 
-    test_loss = evaluate(
+    if args.training_mode == "grpo":
-        model=model,
+        if args.reference_model_path:
-        dataset=test_set,
+            reference_model, _ = load(args.reference_model_path)
-        tokenizer=tokenizer,
+        else:
-        batch_size=args.batch_size,
+            reference_model = model
-        num_batches=args.test_batches,
-        max_seq_length=args.max_seq_length,
-    )
 
-    test_ppl = math.exp(test_loss)
+        test_loss, test_rewards = evaluate_grpo(
+            model=model,
+            reference_model=reference_model,
+            dataset=test_set,
+            tokenizer=tokenizer,
+            batch_size=args.batch_size,
+            num_batches=args.test_batches,
+            max_seq_length=args.max_seq_length,
+            beta=args.beta,
+            group_size=args.group_size,
+            epsilon=args.epsilon,
+            reference_model_path=args.reference_model_path
+        )
+        print(f"Test loss {test_loss:.3f}, Rewards: {test_rewards[0]:.3f}, {test_rewards[1]:.3f}")
+    else:
+        test_loss = evaluate(
+            model=model,
+            dataset=test_set,
+            tokenizer=tokenizer,
+            batch_size=args.batch_size,
+            num_batches=args.test_batches,
+            max_seq_length=args.max_seq_length,
+        )
 
-    print(f"Test loss {test_loss:.3f}, Test ppl {test_ppl:.3f}.")
+        test_ppl = math.exp(test_loss)
+
+        print(f"Test loss {test_loss:.3f}, Test ppl {test_ppl:.3f}.")
 
 
 def run(args, training_callback: TrainingCallback = None):

llms/mlx_lm/tuner/grpo_trainer.py

@@ -22,13 +22,7 @@ generate()
 class GRPOTrainingArgs(TrainingArgs):
     group_size: int = field(
         default=4,
-        metadata={"help": "Number of response sper prompt."},
+        metadata={"help": "Number of responses per prompt."},
-    )
-    is_reference_free: bool = field(
-        default=False,
-        metadata={
-            "help": "Whether to use reference-free DPO training."
-        }
     )
     beta: float = field(
         default=0.1, metadata={"help": "KL penalty coefficient."}