Update backup-precision option description (#1055)

* Update backup-precision option description * Trigger Tests * Trigger Tests
huggingface · Dec 11, 2024 · 41d9a37 · 41d9a37
1 parent 34bf5ae
commit 41d9a37
Show file tree

Hide file tree

Showing 2 changed files with 3 additions and 3 deletions.
diff --git a/docs/source/openvino/export.mdx b/docs/source/openvino/export.mdx
@@ -84,8 +84,8 @@ Optional arguments:
                         The group size to use for quantization. Recommended value is 128 and -1 uses per-column
                         quantization.
   --backup-precision {none,int8_sym,int8_asym}
-                        Defines a backup precision for mixed-precision weight compression. Only valid for int4 weight
-                        format. If not provided, backup precision is int8_asym. 'none' stands for original floating-
+                        Defines a backup precision for mixed-precision weight compression. Only valid for 4-bit weight
+                        formats. If not provided, backup precision is int8_asym. 'none' stands for original floating-
                         point precision of the model weights, in this case weights are retained in their original
                         precision without any quantization. 'int8_sym' stands for 8-bit integer symmetric quantization
                         without zero point. 'int8_asym' stands for 8-bit integer asymmetric quantization with zero

diff --git a/optimum/commands/export/openvino.py b/optimum/commands/export/openvino.py
@@ -123,7 +123,7 @@ def parse_args_openvino(parser: "ArgumentParser"):
         choices=["none", "int8_sym", "int8_asym"],
         default=None,
         help=(
-            "Defines a backup precision for mixed-precision weight compression. Only valid for int4 weight format. "
+            "Defines a backup precision for mixed-precision weight compression. Only valid for 4-bit weight formats. "
             "If not provided, backup precision is int8_asym. 'none' stands for original floating-point precision of "
             "the model weights, in this case weights are retained in their original precision without any "
             "quantization. 'int8_sym' stands for 8-bit integer symmetric quantization without zero point. 'int8_asym' "