diff --git a/src/llamafactory/model/utils/quantization.py b/src/llamafactory/model/utils/quantization.py
index 95412e7c..161ad5aa 100644
--- a/src/llamafactory/model/utils/quantization.py
+++ b/src/llamafactory/model/utils/quantization.py
@@ -131,7 +131,7 @@ def configure_quantization(
                 bnb_4bit_compute_dtype=model_args.compute_dtype,
                 bnb_4bit_use_double_quant=model_args.double_quantization,
                 bnb_4bit_quant_type=model_args.quantization_type,
-                bnb_4bit_quant_storage=model_args.compute_dtype,  # crucial for fsdp qlora
+                bnb_4bit_quant_storage=model_args.compute_dtype,  # crucial for fsdp+qlora
             )
 
         if is_deepspeed_zero3_enabled() or is_fsdp_enabled() or model_args.quantization_device_map == "auto":
@@ -141,6 +141,7 @@ def configure_quantization(
             require_version("transformers>=4.39.0", "To fix: pip install transformers>=4.39.0")
             require_version("accelerate>=0.28.0", "To fix: pip install accelerate>=0.28.0")
             require_version("bitsandbytes>=0.43.0", "To fix: pip install bitsandbytes>=0.43.0")
+            init_kwargs["torch_dtype"] = model_args.compute_dtype  # fsdp+qlora requires same dtype
         else:
             init_kwargs["device_map"] = {"": get_current_device()}