diff --git a/.github/workflows/nv-v100-legacy.yml b/.github/workflows/nv-v100-legacy.yml index 1a35f1f1..346e7f18 100644 --- a/.github/workflows/nv-v100-legacy.yml +++ b/.github/workflows/nv-v100-legacy.yml @@ -36,7 +36,7 @@ jobs: - name: Install dependencies run: | pip install git+https://github.com/microsoft/DeepSpeed.git - pip install git+https://github.com/huggingface/transformers.git@v4.42.4 + pip install git+https://github.com/huggingface/transformers.git@v4.43.4 pip install -U accelerate ds_report diff --git a/mii/legacy/models/load_models.py b/mii/legacy/models/load_models.py index cfbf455f..bbb7e37b 100644 --- a/mii/legacy/models/load_models.py +++ b/mii/legacy/models/load_models.py @@ -75,6 +75,8 @@ def load_models(model_config): if model_config.profile_model_time: engine.profile_model_time() if hasattr(inference_pipeline, "model"): + engine._parameters = inference_pipeline.model._parameters + engine.training = inference_pipeline.model.training inference_pipeline.model = engine elif model_config.enable_zero: