Upload train_lora.py with huggingface_hub

train_lora.py

@@ -113,8 +113,8 @@ def main():
 
     model = AutoModelForCausalLM.from_pretrained(MODEL_NAME, **model_kwargs)
 
-    if USE_4BIT and "deepseek" not in MODEL_NAME.lower():
-        model = prepare_model_for_kbit_training(model)
+    # prepare_model_for_kbit_training removed — peft 0.18+ uses set_submodule
+    # which requires PyTorch 2.5+, but RunPod image has 2.4. SFTTrainer handles it.
 
     # ---- LoRA config ----
     target_modules = get_target_modules(MODEL_NAME)