Segfault in the logOutputs() Method (GraphView.cpp)

I encountered a segfault in the Aidge::GraphView::logOutputs function when the data was on the GPU. The issue was caused by the use of getRawOutput(outIdx), which does not properly handle GPU memory and leads to a crash.

Replaced: fmt::print(fp.get(), "{}\n", nodePtr->getOperator()->getRawOutput(outIdx)->toString().c_str());

With:

std::shared_ptr<Aidge::Tensor> oTensor = std::static_pointer_cast<OperatorTensor>(nodePtr->getOperator())->getOutput(outIdx);
std::shared_ptr<Tensor> fallback;
const Tensor& localTensor = oTensor->refCastFrom(fallback, DataType::Float64, "cpu");
fmt::print(fp.get(), "{}\n", localTensor.toString().c_str());

This ensures that the data is properly cast to CPU memory before being used for logging.

Steps to reproduce: Run logOutputs() on a graph with data stored on the GPU. Before the fix: results in a segfault. After the fix: the function executes correctly.

added Bug 🐛 label

changed the description

Thanks, your fix is correct but you should not convert to float64: logOutputs() should display data as it is.

mentioned in commit f4d3a684

mentioned in merge request !321 (merged)

Will be fixed in !321 (merged).

closed

mentioned in commit 9141f35d

mentioned in commit 6acdc4aa

changed milestone to %aidge v0.5.1

mentioned in commit 451ac9e9

mentioned in commit afeb3f7a