GitAuto: pytorch実行時にエラー（RuntimeError: Attempting to deserialize object on a CUDA device but torch.cuda.is_available() is False. If you are running on a CPU-only machine, please use torch.load with map_location=torch.device('cpu') to map your storages to the CPU.）

[gitauto-ai[bot]]

Resolves #68

Why the bug occurs

The application attempts to deserialize a PyTorch model on a CUDA device when CUDA is not available. This results in a RuntimeError because the model was trained on a GPU, but the current environment does not have CUDA support enabled.

How to reproduce

1. Run the application on a CPU-only machine.
2. The application tries to load a PyTorch model using torch.load without specifying the map_location.
3. A RuntimeError is raised:

   RuntimeError: Attempting to deserialize object on a CUDA device but torch.cuda.is_available() is False. If you are running on a CPU-only machine, please use torch.load with map_location=torch.device('cpu') to map your storages to the CPU.

How to fix

Update the torch.load call in game_manager/machine_learning/block_controller_train_sample.py to include the map_location parameter. This ensures that the model is loaded onto the CPU when CUDA is not available.

Changes to be made:

• Locate the line where the model is loaded:

  self.model = torch.load(predict_weight)

• Modify it to:

  self.model = torch.load(predict_weight, map_location=torch.device('cpu'))

This change ensures compatibility across environments with or without CUDA support.

Test these changes locally

git checkout -b gitauto/issue-68-caa2b879-c1b8-4048-a32d-c88f276000ee
git pull origin gitauto/issue-68-caa2b879-c1b8-4048-a32d-c88f276000ee