commit 11404c21dc0b5409e85686c426c9ae7c20944147
Author: Daniel Bevenius <daniel.bevenius@gmail.com>
Date:   Tue Apr 28 08:53:45 2026 +0200

    convert : avoid nvfp4 processing for mmproj model
    
    This commit enables avoiding nvfp4 processing for mmproj models as the
    test language model does not need to be processed for these models and
    they also don't contain the mapping of the text model tensors which will
    cause errors during conversion.

diff --git a/convert_hf_to_gguf.py b/convert_hf_to_gguf.py
index f5796cb5d..03aa957f0 100755
--- a/convert_hf_to_gguf.py
+++ b/convert_hf_to_gguf.py
@@ -728,6 +728,9 @@ class ModelBase:
 
         del experts, merged
 
+    def _needs_nvfp4_processing(self) -> bool:
+        return True
+
     def prepare_tensors(self):
         # detect NVFP4 quantization (ModelOpt format)
         quant_algo = (self.hparams.get("quantization_config") or {}).get("quant_algo")
@@ -758,7 +761,7 @@ class ModelBase:
         # NVFP4 weights are repacked and written directly to gguf_writer.
         # This must run before dequant_model so NVFP4 tensors are removed
         # from model_tensors, leaving only non-NVFP4 (e.g. FP8) for dequant.
-        if self._is_nvfp4:
+        if self._is_nvfp4 and self._needs_nvfp4_processing():
             self._generate_nvfp4_tensors()
 
         self.dequant_model()
@@ -2190,6 +2193,10 @@ class MmprojModel(ModelBase):
                 # merge configs
                 self.preprocessor_config = {**self.preprocessor_config, **cfg}
 
+    def _needs_nvfp4_processing(self) -> bool:
+        # nvfp4 quantization applies to the text model only.
+        return False
+
     def get_vision_config(self) -> dict[str, Any] | None:
         config_name = "vision_config" if not self.is_mistral_format else "vision_encoder"
         return self.global_config.get(config_name)
@@ -4450,6 +4457,12 @@ class NemotronNanoV2VLModel(MmprojModel):
         }
         return vision_config
 
+    def dequant_model(self):
+        if self._is_nvfp4:
+            # Skip nvfp4 quantization for vision/audio model.
+            return
+        super().dequant_model()
+
     def set_gguf_parameters(self):
         if "image_mean" not in self.preprocessor_config:
             self.preprocessor_config["image_mean"] = [0.485, 0.456, 0.406]

INFO:hf-to-gguf:Exporting model...
Traceback (most recent call last):
  File "/home/danbev/work/llama.cpp/examples/model-conversion/../../convert_hf_to_gguf.py", line 13586, in <module>
    main()
  File "/home/danbev/work/llama.cpp/examples/model-conversion/../../convert_hf_to_gguf.py", line 13580, in main
    model_instance.write()
  File "/home/danbev/work/llama.cpp/examples/model-conversion/../../convert_hf_to_gguf.py", line 933, in write
    self.prepare_tensors()
  File "/home/danbev/work/llama.cpp/examples/model-conversion/../../convert_hf_to_gguf.py", line 775, in prepare_tensors
    for name, data_torch in chain(self.generate_extra_tensors(), self.get_tensors()):
  File "/home/danbev/work/llama.cpp/examples/model-conversion/../../convert_hf_to_gguf.py", line 527, in get_tensors
    yield name, gen()
                ^^^^^
  File "/home/danbev/work/llama.cpp/examples/model-conversion/../../convert_hf_to_gguf.py", line 511, in <lambda>
    self.model_tensors[weight_name] = lambda w=w, s=s: dequant_simple(w(), s(), None)
                                                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/danbev/work/llama.cpp/examples/model-conversion/../../convert_hf_to_gguf.py", line 328, in dequant_simple
    return weight.float() * scale
           ~~~~~~~~~~~~~~~^~~~~~~
  File "/home/danbev/work/llama.cpp/examples/model-conversion/../../gguf-py/gguf/lazy.py", line 40, in wrapped_special_op
    return type(self)._wrap_fn(
           ^^^^^^^^^^^^^^^^^^^^
  File "/home/danbev/work/llama.cpp/examples/model-conversion/../../gguf-py/gguf/lazy.py", line 126, in wrapped_fn
    res = fn(*meta_args, **kwargs)
          ^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/danbev/work/llama.cpp/venv/lib/python3.12/site-packages/torch/_prims_common/wrappers.py", line 291, in _fn
    result = fn(*args, **kwargs)
             ^^^^^^^^^^^^^^^^^^^
  File "/home/danbev/work/llama.cpp/venv/lib/python3.12/site-packages/torch/_prims_common/wrappers.py", line 143, in _fn
    result = fn(**bound.arguments)
             ^^^^^^^^^^^^^^^^^^^^^
  File "/home/danbev/work/llama.cpp/venv/lib/python3.12/site-packages/torch/_refs/__init__.py", line 1095, in _ref
    a, b = _maybe_broadcast(a, b)
           ^^^^^^^^^^^^^^^^^^^^^^
  File "/home/danbev/work/llama.cpp/venv/lib/python3.12/site-packages/torch/_refs/__init__.py", line 437, in _maybe_broadcast
    common_shape = _broadcast_shapes(
                   ^^^^^^^^^^^^^^^^^^
  File "/home/danbev/work/llama.cpp/venv/lib/python3.12/site-packages/torch/_refs/__init__.py", line 425, in _broadcast_shapes
    torch._check(
  File "/home/danbev/work/llama.cpp/venv/lib/python3.12/site-packages/torch/__init__.py", line 1656, in _check
    _check_with(RuntimeError, cond, message)
  File "/home/danbev/work/llama.cpp/venv/lib/python3.12/site-packages/torch/__init__.py", line 1638, in _check_with
    raise error_type(message_evaluated)
RuntimeError: Attempting to broadcast a dimension of length 116 at -1! Mismatching argument at index 1 had torch.Size([2688, 116]); but expected shape should be broadcastable to [2688, 928]