Upload transformers-bitsandbytes.patch

unsloth puzzles : fix Q4 model loading error in official notebook :
requires installing HF transformers module locally after patching

transformers-bitsandbytes.patch

@@ -0,0 +1,13 @@
+diff --git a/src/transformers/utils/import_utils.py b/src/transformers/utils/import_utils.py
+index aa7be764c..c1f7ccadc 100755
+--- a/src/transformers/utils/import_utils.py
++++ b/src/transformers/utils/import_utils.py
+@@ -921,7 +921,7 @@ def is_torch_xpu_available(check_device=False):
+ 
+ @lru_cache()
+ def is_bitsandbytes_available():
+-    if not is_torch_available() or not _bitsandbytes_available:
++    if not is_torch_available():
+         return False
+ 
+     import torch