fix cpu and xpu issue (#2116)

Signed-off-by: Wang, Yi A <[email protected]>
huggingface · Jun 25, 2024 · e563983 · e563983
1 parent 9e2fdf5
commit e563983
Show file tree

Hide file tree

Showing 7 changed files with 16 additions and 7 deletions.
diff --git a/server/text_generation_server/models/flash_causal_lm.py b/server/text_generation_server/models/flash_causal_lm.py
@@ -768,7 +768,10 @@ def init_kv_cache(
         empty_cache()
 
         element_size = torch.tensor([], dtype=dtype).element_size()
-        x = BLOCK_SIZE // element_size
+        if SYSTEM == "ipex" and device.type == "xpu":
+            x = 1
+        else:
+            x = BLOCK_SIZE // element_size
 
         if SYSTEM == "ipex" and device == torch.device("cpu"):
             self.kv_cache = [

diff --git a/server/text_generation_server/models/flash_gpt2.py b/server/text_generation_server/models/flash_gpt2.py
@@ -37,9 +37,10 @@ def __init__(
         elif SYSTEM == "ipex":
             if hasattr(torch, "xpu") and torch.xpu.is_available():
                 device = torch.device(f"xpu:{rank}")
+                dtype = torch.float16 if dtype is None else dtype
             else:
                 device = torch.device("cpu")
-            dtype = torch.float16 if dtype is None else dtype
+                dtype = torch.bfloat16 if dtype is None else dtype
         else:
             raise NotImplementedError("FlashGPT2 is only available on GPU")
 

diff --git a/server/text_generation_server/models/flash_llama.py b/server/text_generation_server/models/flash_llama.py
@@ -37,9 +37,10 @@ def __init__(
         elif SYSTEM == "ipex":
             if hasattr(torch, "xpu") and torch.xpu.is_available():
                 device = torch.device(f"xpu:{rank}")
+                dtype = torch.float16 if dtype is None else dtype
             else:
                 device = torch.device("cpu")
-            dtype = torch.float16 if dtype is None else dtype
+                dtype = torch.bfloat16 if dtype is None else dtype
         else:
             raise NotImplementedError("FlashLlama is only available on GPU")
 

diff --git a/server/text_generation_server/models/flash_mistral.py b/server/text_generation_server/models/flash_mistral.py
@@ -41,9 +41,10 @@ def __init__(
         elif SYSTEM == "ipex":
             if hasattr(torch, "xpu") and torch.xpu.is_available():
                 device = torch.device(f"xpu:{rank}")
+                dtype = torch.float16 if dtype is None else dtype
             else:
                 device = torch.device("cpu")
-            dtype = torch.float16 if dtype is None else dtype
+                dtype = torch.bfloat16 if dtype is None else dtype
         else:
             raise NotImplementedError("FlashMistral is only available on GPU")
 

diff --git a/server/text_generation_server/models/flash_neox.py b/server/text_generation_server/models/flash_neox.py
@@ -36,9 +36,10 @@ def __init__(
         elif SYSTEM == "ipex":
             if hasattr(torch, "xpu") and torch.xpu.is_available():
                 device = torch.device(f"xpu:{rank}")
+                dtype = torch.float16 if dtype is None else dtype
             else:
                 device = torch.device("cpu")
-            dtype = torch.float16 if dtype is None else dtype
+                dtype = torch.bfloat16 if dtype is None else dtype
         else:
             raise NotImplementedError("FlashNeoX is only available on GPU")
 

diff --git a/server/text_generation_server/models/flash_rw.py b/server/text_generation_server/models/flash_rw.py
@@ -37,9 +37,10 @@ def __init__(
         elif SYSTEM == "ipex":
             if hasattr(torch, "xpu") and torch.xpu.is_available():
                 device = torch.device(f"xpu:{rank}")
+                dtype = torch.float16 if dtype is None else dtype
             else:
                 device = torch.device("cpu")
-            dtype = torch.float16 if dtype is None else dtype
+                dtype = torch.bfloat16 if dtype is None else dtype
         else:
             raise NotImplementedError("FlashRW is only available on GPU")
 

diff --git a/server/text_generation_server/models/flash_santacoder.py b/server/text_generation_server/models/flash_santacoder.py
@@ -40,9 +40,10 @@ def __init__(
         elif SYSTEM == "ipex":
             if hasattr(torch, "xpu") and torch.xpu.is_available():
                 device = torch.device(f"xpu:{rank}")
+                dtype = torch.float16 if dtype is None else dtype
             else:
                 device = torch.device("cpu")
-            dtype = torch.float16 if dtype is None else dtype
+                dtype = torch.bfloat16 if dtype is None else dtype
         else:
             raise NotImplementedError("FlashSantacoderSharded is only available on GPU")