From b209b8f7b683f65afef659088e71c34ad745de47 Mon Sep 17 00:00:00 2001
From: Yina Chen <33650826+cyita@users.noreply.github.com>
Date: Thu, 7 Sep 2023 23:56:36 +0800
Subject: [PATCH] [LLM] Fix arc qtype != q4_0 generate issue (#8920)

* Fix arc precision!=q4_0 generate issue

* meet comments
---
 .../llm/src/bigdl/llm/transformers/low_bit_linear.py   | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/python/llm/src/bigdl/llm/transformers/low_bit_linear.py b/python/llm/src/bigdl/llm/transformers/low_bit_linear.py
index 241f6109..2a040bf0 100644
--- a/python/llm/src/bigdl/llm/transformers/low_bit_linear.py
+++ b/python/llm/src/bigdl/llm/transformers/low_bit_linear.py
@@ -98,7 +98,10 @@ def ggml_q_format_convet_cpu2xpu(tensor: torch.Tensor, num_elem: int, qtype: int
 
     src = ctypes.c_void_p(tensor.data.data_ptr())
 
-    dst_tensor = torch.empty_like(tensor)
+    if qtype == ggml_tensor_qtype["sym_int4"]:
+        dst_tensor = torch.empty_like(tensor)
+    else:
+        return tensor
     dst = ctypes.c_void_p(dst_tensor.data.data_ptr())
     ggml.ggml_q_format_convet_cpu2xpu(src, dst, num_elem, qtype)
     return dst_tensor
@@ -114,7 +117,10 @@ def ggml_q_format_convet_xpu2cpu(tensor: torch.Tensor, num_elem: int, qtype: int
 
     src = ctypes.c_void_p(tensor.data.data_ptr())
 
-    dst_tensor = torch.empty_like(tensor)
+    if qtype == ggml_tensor_qtype["sym_int4"]:
+        dst_tensor = torch.empty_like(tensor)
+    else:
+        return tensor
     dst = ctypes.c_void_p(dst_tensor.data.data_ptr())
     ggml.ggml_q_format_convet_xpu2cpu(src, dst, num_elem, qtype)
     return dst_tensor