macophy

> We just got MOTA of 73.1 70.1 63.5，can not reproduce your results MOTA of 74.8 70.8 63.7 in README.md. > Please give more details. I also got MOTA of...

> I download the pre-trained model and test on the MOT-16 training set, only get the result with MOTA 73.1. How can I get 74.8 MOTA described in readme.md? Does...

我做了一些实验证明，加了加了L1范式后，如@ZongshenXie所说可以使γ更趋近于0（也就是数值较小的γ更多了），但是并不可以使γ值变为0。

> > 我做了一些实验证明，加了加了L1范式后，如@ZongshenXie所说可以使γ更趋近于0（也就是数值较小的γ更多了），但是并不可以使γ值变为0。 > > 代码里面实际上并不是L1，而是一个与γ正负相同的固定值，应该算是L0。另外，剪枝并不需要使γ严格等于0，而是让部分γ足够小，也就是让这部分channel对网络的影响变小，从而剪枝的时候把它们去掉后网络依然能够正常运行 L0的说法还是不太严谨的，这里援引netslim原作者的回答https://github.com/Eric-mingjie/network-slimming/issues/31

> decrease batchsize, try 1 @juanmanuelrq Great! I meet the same problem. Solved!

> The first cell automatically uninstalls NumPy 1.25.2 and installs NumPy 1.26.4, so you don't need to install numpy again. When I run first two cells, it does throw the...