h6197627

RAM usage regeression between v1.0.1 and v2.8.1

1

I am in the process of upgrading from MNN v1.0.1 to v2.8.1 in my project. While doing initial research I noticed that one model (ResNet100) requires x2 RAM in v2.8.1...

There is inconsistency between outputs produced by MNN and ONNXRuntime for Deconvolution op with manually specified output shape [ONNX model](https://github.com/user-attachments/files/23168308/deconv.onnx.zip) (single ConvTranspose layer) [MNN model](https://github.com/user-attachments/files/23168316/deconv.mnn.zip) (converted from ONNX model with...

## Description There is inconsistency between outputs produced by TensorRT and ONNXRuntime for Deconvolution op with manually specified output shape [ONNX model](https://github.com/user-attachments/files/23215529/model.onnx.zip) (single ConvTranspose layer) [Compare python script](https://github.com/user-attachments/files/23215552/cmp.py), that load...