fix(trainer): 修复参数传递问题和检查点保存问题

khaosz/config/train_config.py

@@ -93,4 +93,13 @@ class TrainConfig:
     kwargs: dict = field(
         default_factory=dict,
         metadata={"help": "Other arguments."}
-    )
+    )
+    
+    def __post_init__(self):
+        self.validate()
+    
+    def validate(self):
+        required_fields = ["model", "strategy", "dataset", "optimizer", "scheduler"]
+        for field_name in required_fields:
+            if getattr(self, field_name) is None:
+                raise ValueError(f"{field_name} is required.")

khaosz/trainer/train_callback.py

@@ -94,18 +94,17 @@ class CheckpointCallback(TrainCallback):
     def __init__(self, interval: int, save_dir: str):
         self.interval = interval
         self.save_dir = save_dir
-        self.checkpoint = None
         self.last_ckpt_iter = 0
     
     def _save_checkpoint(self, context: 'TrainContext'):
         save_path = os.path.join(self.save_dir, f"epoch_{context.epoch}iter_{context.iteration}")
-        self.checkpoint = Checkpoint(
+        context.checkpoint = Checkpoint(
             context.optimizer.state_dict(), 
             context.scheduler.state_dict(), 
             context.epoch, 
             context.iteration
         )
-        self.checkpoint.save(save_path)
+        context.checkpoint.save(save_path)
         self.last_ckpt_iter = context.iteration
     
     def on_batch_end(self, context: 'TrainContext'):

tests/test_train_config.py

@@ -11,10 +11,18 @@ def test_different_batch_sizes(base_test_env, random_dataset):
     batch_sizes = [1, 2, 4, 8]
     
     for batch_size in batch_sizes:
+        schedule_config = CosineScheduleConfig(
+            warmup_steps=10,
+            total_steps=20
+        )
         optimizer = torch.optim.AdamW(base_test_env["model"].parameters())
+        scheduler = SchedulerFactory.load(optimizer, schedule_config)
         train_config = TrainConfig(
+            strategy="seq",
+            model=base_test_env["model"],
             dataset=random_dataset,
             optimizer=optimizer,
+            scheduler=scheduler,
             checkpoint_dir=base_test_env["test_dir"],
             n_epoch=1,
             batch_size=batch_size,
@@ -67,10 +75,18 @@ def test_memory_efficient_training(base_test_env, random_dataset):
     ]
     
     for config in small_batch_configs:
+        schedule_config = CosineScheduleConfig(
+            warmup_steps=10,
+            total_steps=20
+        )
         optimizer = torch.optim.AdamW(base_test_env["model"].parameters())
+        scheduler = SchedulerFactory.load(optimizer, schedule_config)
         train_config = TrainConfig(
+            strategy="seq",
+            model=base_test_env["model"],
             dataset=random_dataset,
             optimizer=optimizer,
+            scheduler=scheduler,
             checkpoint_dir=base_test_env["test_dir"],
             n_epoch=1,
             batch_size=config["batch_size"],