TabularModel.load_model issue: Trained GPU model and loading CPU-only machine

[arcturus9725]

Hello,
I am trying to load model and predict in CPU only machine,
which is trained with GPU machine.

I tried following code:
tabular_model = TabularModel.load_model("md_tmp2", map_location=torch.device('cpu'))

But error comes out:
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.

Please let me know how to deal with this.

Desktop:

• OS: Windows
• Version

• pytorch_tabular 1.1.1
• torch 2.5.1+cu124(GPU Machine), torch 2.5.1(CPU Only machine)

[dosubot]

Hey @arcturus9725! 👋 I'm here to help you with your code issues and questions. Let's squash those bugs together!

To resolve the issue of loading a model trained on a GPU machine and predicting on a CPU-only machine, you should ensure that the map_location argument is correctly set to 'cpu' when using TabularModel.load_model. Here's how you can do it:

from pytorch_tabular import TabularModel

# Load the model with map_location set to 'cpu'
tabular_model = TabularModel.load_model("md_tmp2", map_location='cpu')

By setting map_location='cpu', the model will be loaded onto the CPU, which should prevent the error you're encountering ^[1].

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Other}

[arcturus9725]

Hello, I tried your code but same error came out.

from pytorch_tabular import TabularModel

tabular_model = TabularModel.load_model("md_tmp2", map_location='cpu')

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.

Thanks.