Upload train_glm47_flash.py with huggingface_hub

train_glm47_flash.py

@@ -158,7 +158,7 @@ training_config = SFTConfig(
     per_device_eval_batch_size=1,
     gradient_accumulation_steps=16,  # Effective batch size: 16
     learning_rate=2e-4,
-    max_seq_length=1024,  # Reduced for memory
+    max_length=1024,  # Reduced for memory
 
     # Memory optimization
     gradient_checkpointing=True,