Cleanup

modules/api/models.py

@@ -105,7 +105,7 @@ class FaceIDItem(BaseModel):
     structure: float = Field(title="Structure", default=1, gt=0, le=1, description="Structure to use, must be between 0 and 1.")
     rank: float = Field(title="Rank", default=128, ge=4, le=256, description="Rank to use, must be between 4 and 256.")
     override_sampler: bool = Field(title="Override Sampler", default=True, description="Should the sampler be overriden?")
-    tokens: int = Field("Tokens", default=4, ge=1, le=16, description="Amount of tokens to use, must be between 1 and 16.")
+    tokens: int = Field(title="Tokens", default=4, ge=1, le=16, description="Amount of tokens to use, must be between 1 and 16.")
     cache_model: bool = Field(title="Cache", default=True, description="Should the model be cached?")
 
 

modules/intel/ipex/attention.py

@@ -124,7 +124,7 @@ def torch_bmm_32_bit(input, mat2, *, out=None):
                 )
     else:
         return original_torch_bmm(input, mat2, out=out)
-    torch.xpu.synchronize()
+    torch.xpu.synchronize(input.device)
     return hidden_states
 
 original_scaled_dot_product_attention = torch.nn.functional.scaled_dot_product_attention
@@ -173,5 +173,5 @@ def scaled_dot_product_attention_32_bit(query, key, value, attn_mask=None, dropo
                 )
     else:
         return original_scaled_dot_product_attention(query, key, value, attn_mask=attn_mask, dropout_p=dropout_p, is_causal=is_causal)
-    torch.xpu.synchronize()
+    torch.xpu.synchronize(query.device)
     return hidden_states

modules/intel/ipex/diffusers.py

@@ -149,7 +149,7 @@ class SlicedAttnProcessor: # pylint: disable=too-few-public-methods
 
                         hidden_states[start_idx:end_idx, start_idx_2:end_idx_2] = attn_slice
                         del attn_slice
-                torch.xpu.synchronize()
+                torch.xpu.synchronize(query.device)
             else:
                 query_slice = query[start_idx:end_idx]
                 key_slice = key[start_idx:end_idx]
@@ -284,7 +284,7 @@ class AttnProcessor:
 
                     hidden_states[start_idx:end_idx] = attn_slice
                     del attn_slice
-            torch.xpu.synchronize()
+            torch.xpu.synchronize(query.device)
         else:
             attention_probs = attn.get_attention_scores(query, key, attention_mask)
             hidden_states = torch.bmm(attention_probs, value)