Workaround xformers bug.

comfy/ldm/modules/attention.py

@@ -297,8 +297,8 @@ def attention_split(q, k, v, heads, mask=None):
 BROKEN_XFORMERS = False
 try:
     x_vers = xformers.__version__
-    #I think 0.0.23 is also broken (q with bs bigger than 65535 gives CUDA error)
+    # XFormers bug confirmed on all versions from 0.0.21 to 0.0.26 (q with bs bigger than 65535 gives CUDA error)
-    BROKEN_XFORMERS = x_vers.startswith("0.0.21") or x_vers.startswith("0.0.22") or x_vers.startswith("0.0.23")
+    BROKEN_XFORMERS = x_vers.startswith("0.0.2") and not x_vers.startswith("0.0.20")
 except:
     pass