small change (#12439)

python/llm/src/ipex_llm/transformers/convert.py

@@ -1248,7 +1248,7 @@ def _optimize_post(model, lightweight_bmm=False):
     try:
         from diffusers import DiffusionPipeline
         if isinstance(model, DiffusionPipeline):
-            from ipex_llm.transformers.models.sd15 import AttnProcessor2_0
+            from ipex_llm.transformers.models.sd import AttnProcessor2_0
             model.unet.set_attn_processor(AttnProcessor2_0())
             return model
     except ModuleNotFoundError:

python/llm/src/ipex_llm/transformers/models/sd.py

@@ -106,8 +106,8 @@ class AttnProcessor2_0:
 
         # the output of sdp = (batch, num_heads, seq_len, head_dim)
         # IPEX-LLM changes start
-        # padding head_dim 40 to 64
         if query.device.type == "xpu" and query.dtype in [torch.half, torch.float]:
+            # padding head_dim 40 to 64
             query, key, value = padding_qkv_hd(query, key, value, 40, 64)
 
             if use_sdp_non_causal(head_dim, query.device, query.dtype):

python/llm/src/ipex_llm/transformers/models/utils.py

@@ -329,7 +329,7 @@ def use_sdp_causal(q_len, kv_len, head_dim, query_states, training):
 
 def use_sdp_non_causal(head_dim, device, dtype):
     return (
-        head_dim in [40, 64, 80]
+        head_dim in [64, 80, 128]
         and device.type == "xpu"                # GPU
         and dtype in [torch.float, torch.half]  # fp32/fp16
     )