fix memory for online fp8 quantization with streaming weight load (#31914)

Signed-off-by: vasiliy <vasiliy@fb.com>
This commit is contained in:
Vasiliy Kuznetsov
2026-02-02 14:17:42 -05:00
committed by GitHub
parent 5d1aef3004
commit 0130223bd9
5 changed files with 283 additions and 38 deletions
@@ -25,4 +25,4 @@ class DummyModelLoader(BaseModelLoader):
def load_weights(self, model: nn.Module, model_config: ModelConfig) -> None:
# NOTE(woosuk): For accurate performance evaluation, we assign
# random values to the weights.
initialize_dummy_weights(model)
initialize_dummy_weights(model, model_config)