Avoid silent weights corruption when loading Nemotron Nano VL with reusable-buffer loaders like runai distributed streaming (#42244)

Signed-off-by: Noa Neria <nneria@nvidia.com>
This commit is contained in:
Noa Neria
2026-05-11 12:03:14 +00:00
committed by GitHub
parent 617239b70c
commit ac062147fa
2 changed files with 54 additions and 27 deletions
@@ -53,6 +53,19 @@ class _VisionModel:
self.loaded_weights = list(weights)
class _FakeTensor:
"""Sentinel stand-in for torch.Tensor in load_weights tests. Supports the
.detach().clone() chain used by load_weights for buffered mm weights;
both methods return self so identity (and the existing equality
assertions) are preserved through cloning."""
def detach(self):
return self
def clone(self):
return self
def test_nano_nemotron_vl_skips_multimodal_weights_in_text_only_mode():
model = object.__new__(NemotronH_Nano_VL_V2)
language_model = _LanguageModel()
@@ -86,7 +99,7 @@ def test_nano_nemotron_vl_loads_vision_weights_without_sound_encoder():
object.__setattr__(model, "sound_encoder", None)
language_weight = object()
vision_weight = object()
vision_weight = _FakeTensor()
model.load_weights(
[
("language_model.layers.0.weight", language_weight),