nanjoin

generate train data for MPII

3

Someone can answer that for me in the [hyperpose/dataset/mpii_dataset/Dataset/mpii_dataset/generate.py](https://github.com/tensorlayer/hyperpose/blob/master/hyperpose/Dataset/mpii_dataset/generate.py) ``` target_list=[] for kpts,head_bbx in zip(kpts_list,bbx_list): bbx=np.array(head_bbx).copy() bbx[:,2]=bbx[:,2]*4 bbx[:,3]=bbx[:,3]*4 target_list.append({ "kpt":kpts, "mask":None, "bbx":bbx, "head_bbx":head_bbx, "labeled":1 }) ``` What does **bbx[:,2]=bbx[:,2]\*4** and...

量化后的模型在推理的时候net.opt.use_packing_layout的开关会导致结果不一致

2

使用【ncnn20220729】版本，在linux下量化，linux和windows平台推理均出现以下情况： 模型在量化后net.opt.use_packing_layout默认开启的时候，在量化数据集上结果acc只有0.14的 各种调整数据集以及量化的method后，量化数据集的acc连0.3都达不到。 但是将net.opt.use_packing_layout=False后，acc就正常了。 附件是int8的param，模型的输入是112*112*3的输入。麻烦up帮忙看一下，谢谢！@nihui [param.zip](https://github.com/Tencent/ncnn/files/9367505/param.zip)

Fix Incorrect output for A or B with dim=1 in GEMM

1

Fixed a bug in the GEMM layer where incorrect shapes of M, N, and K occurred when matrix A or matrix B had dim=1, leading to incorrect output results