w1005444804

inferenc cost time (2048x600x3) < (2048x512x3)

2

## error log | 日志或报错信息 | ログ ## context | 编译/运行环境 | バックグラウンド win32 ncnn build myself, create my net using interface create_layer("") one by one. ## how to reproduce...

## Description ## Environment **TensorRT Version**: **NVIDIA GPU**: **NVIDIA Driver Version**: **CUDA Version**: **CUDNN Version**: **Operating System**: **Python Version (if applicable)**: **Tensorflow Version (if applicable)**: **PyTorch Version (if applicable)**: **Baremetal...

how to use the input masks??

6

As the title says, when I followed the instructions, I didn't meet my expectations。 instructions： mask_input (np.ndarray): A low resolution mask input to the model, typically coming from a previous...

Is dnnl::convolution_backward_data running on a single core？

9

cpu: i3-8100, 4/4; windows10 just using single convolution, forward and backward. Its speed is four times slower than Pytorch。 Is multi-core not enabled? How to accelerate my dnnl-code？

I just want to use VIT with fp16 precision ， but [W]Running layernorm after self-attention in FP16 may cause overflow. Now I hope other layers run in FP16,but layernorm layers...

std::vector net;//手动创建的net。net是串联的 std::vector blob_mats;//用于net的输入与输出 ncnn::Option opt; bool infer(const ncnn::Mat& input, ncnn::Mat& output) { blob_mats[0] = input; for (size_t i = 0; i < net.size(); i++) { net[i].first->forward(blob_mats[i], blob_mats[i+1], net[i].second); }...

通过Ollama下载了的模型，如何在airllm中直接使用呢

2

期待。。。。 

Is there only one input shape for vit to correctly output mask？

7

If the input of VIt is not 1024x1024 but something else, such as 1024x512 or 768x512, can sam also accurately output the mask

Thank you very much for sharing this project, Could you please offer some advice or guidance on how to approach reading your code?