Delegate ModelExpress loading to package (#24723)
Signed-off-by: Zheng Luo <zheluo@nvidia.com>
This commit is contained in:
@@ -123,6 +123,24 @@ class TestRunaiModelStreamerLoader(CustomTestCase):
|
||||
|
||||
self.assertIsInstance(model_loader, loader_mod.RunaiModelStreamerLoader)
|
||||
|
||||
def test_get_model_loader_uses_remote_instance_for_prequantized_modelopt(self):
|
||||
load_config = LoadConfig(
|
||||
load_format=LoadFormat.REMOTE_INSTANCE,
|
||||
model_loader_extra_config={},
|
||||
)
|
||||
model_config = cast(
|
||||
ModelConfig,
|
||||
SimpleNamespace(
|
||||
quantization="modelopt_fp4",
|
||||
modelopt_quant=False,
|
||||
_is_already_quantized=lambda: True,
|
||||
),
|
||||
)
|
||||
|
||||
model_loader = loader_mod.get_model_loader(load_config, model_config)
|
||||
|
||||
self.assertIsInstance(model_loader, loader_mod.RemoteInstanceModelLoader)
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
unittest.main()
|
||||
|
||||
Reference in New Issue
Block a user