[ { "function_name": "simpo_loss", "file_path": "SimPO/scripts/simpo_loss_calc.py", "function_calls": [ { "function_call_name": "to", "functionality": "Moves the input tensor to the specified device (e.g., CPU, GPU).", "input": "tensor (torch.FloatTensor), device (accelerator_device)", "output": "tensor (torch.FloatTensor)", "location_or_namespace": "torch" }, { "function_call_name": "logsigmoid", "functionality": "Applies the logarithm of the sigmoid function to the input tensor.", "input": "input (torch.Tensor)", "output": "output (torch.Tensor)", "location_or_namespace": "torch.nn.functional" }, { "function_call_name": "relu", "functionality": "Applies the Rectified Linear Unit (ReLU) function element-wise. Specifically, `relu(x) = max(0, x)`.", "input": "input (torch.Tensor)", "output": "output (torch.Tensor)", "location_or_namespace": "torch" }, { "function_call_name": "detach", "functionality": "Returns a new tensor, detached from the current graph. The result will never require gradient.", "input": "tensor (torch.FloatTensor)", "output": "tensor (torch.FloatTensor)", "location_or_namespace": "torch" } ] } ]