From 4fcd87cf7cbb131b3e28e121b29cc588e460eb40 Mon Sep 17 00:00:00 2001
From: Aleksei Nikiforov
 <103434461+AlekseiNikiforovIBM@users.noreply.github.com>
Date: Thu, 27 Nov 2025 11:35:38 +0100
Subject: [PATCH] gguf-py : skip endian-conversion of MXFP4 data (#17523)

* gguf_convert_endian.py: skip MXFP4 data

* Use gguf.constants.GGML_QUANT_SIZES to determine block sizes
---
 gguf-py/gguf/scripts/gguf_convert_endian.py | 30 +++++++++------------
 1 file changed, 12 insertions(+), 18 deletions(-)

diff --git a/gguf-py/gguf/scripts/gguf_convert_endian.py b/gguf-py/gguf/scripts/gguf_convert_endian.py
index 0bda490a20..86bf87846c 100755
--- a/gguf-py/gguf/scripts/gguf_convert_endian.py
+++ b/gguf-py/gguf/scripts/gguf_convert_endian.py
@@ -19,6 +19,11 @@ import gguf
 logger = logging.getLogger("gguf-convert-endian")
 
 
+def byteswap_noop(tensor, block_offs):
+    # this function is used when byteswapping is not needed
+    pass
+
+
 def byteswap_q4_0(tensor, block_offs):
     # Each block_q4_0 consists of an f16 delta (scaling factor) followed by 16 int8 quantizations.
 
@@ -55,22 +60,11 @@ def byteswap_q6_k(tensor, block_offs):
 
 
 byteswap_tensors = {
-    gguf.GGMLQuantizationType.Q4_0: {
-        "block_size": 18, # 18 bytes = <f16 delta scaling factor> + 16 * <int8 quant>
-        "byteswap_func": byteswap_q4_0,
-    },
-    gguf.GGMLQuantizationType.Q8_0: {
-        "block_size": 34, # 34 bytes = <f16 delta scaling factor> + 32 * <int8 quant>
-        "byteswap_func": byteswap_q8_0,
-    },
-    gguf.GGMLQuantizationType.Q4_K: {
-        "block_size": 144, # 144 bytes = 2 * <f16 delta scaling factor> + 140 * <int8 quant>
-        "byteswap_func": byteswap_q4_k,
-    },
-    gguf.GGMLQuantizationType.Q6_K: {
-        "block_size": 210, # 210 bytes = <f16 delta scaling factor> + 208 * <int8 quant>
-        "byteswap_func": byteswap_q6_k,
-    },
+    gguf.GGMLQuantizationType.Q4_0:  byteswap_q4_0,
+    gguf.GGMLQuantizationType.Q8_0:  byteswap_q8_0,
+    gguf.GGMLQuantizationType.Q4_K:  byteswap_q4_k,
+    gguf.GGMLQuantizationType.Q6_K:  byteswap_q6_k,
+    gguf.GGMLQuantizationType.MXFP4: byteswap_noop,
 }
 
 
@@ -135,8 +129,8 @@ def convert_byteorder(reader: gguf.GGUFReader, args: argparse.Namespace) -> None
 
             tensor.data.resize(newshape)
 
-            block_size    = byteswap_tensors[tensor.tensor_type]["block_size"]
-            byteswap_func = byteswap_tensors[tensor.tensor_type]["byteswap_func"]
+            block_size    = gguf.constants.GGML_QUANT_SIZES[tensor.tensor_type][1]
+            byteswap_func = byteswap_tensors[tensor.tensor_type]
 
             n_blocks = len(tensor.data) // block_size
             for block_num in (inner_pbar := tqdm(range(n_blocks), desc="Byte-swapping Blocks", leave=False)):