Sanqing Yuan

Results 6 issues of Sanqing Yuan

用的是默认的配置:6x6 board and 4 in a row. macos上跑的。 batch i:1100, episode_len:21 kl:0.00058,lr_multiplier:11.391,loss:4.518421649932861,entropy:3.5188446044921875,explained_var_old:0.000,explained_var_new:0.000 current self-play batch: 1100 num_playouts:1000, win: 2, lose: 8, tie:0 请指教。

使用3角麦的数据,DOA容易在正确方向和对角线方向间跳跃,比如正确方向是60度,声源位置不变的情况下会时不时指向240度;也就是容易在角度a和(a+180)间切换,请问是什么原因?致谢!

在android studio使用模拟器实测,用的 [The squeezencnn example](https://github.com/Tencent/ncnn/tree/master/examples/squeezencnn),测试了API 24 到API 30,使用ncnn-20220216-android-vulkan,gpu_count通过ncnn::get_gpu_count()获得,结果如下: ``` Android API gpu_count 7 24 0 7.1.1 25 0 8 26 0 8.1 27 0 9 28 0 10 29 1...

Instead of ![屏幕截图 2021-02-19 184109](https://user-images.githubusercontent.com/9622783/108494141-28588f80-72e2-11eb-92f5-5b8f8ebd3fb1.png) can I get something like this? ![屏幕截图 2021-02-19 183822](https://user-images.githubusercontent.com/9622783/108494191-36a6ab80-72e2-11eb-9adc-0c3822cef23d.png)

Hi thanks much for sharing. I used only pump data for training and testing because my computer is slow. After 10 epoch training, auc=77.4, but after 300 epoch, auc is...

出错 GWPE (line 106) enhanced_frames = irfft(spec, fft_len, 2); 我改为ifft后可以跑了,但是输出声音不正常。我用的是单声道,这个是不是只能用阵列啊。 请帮忙看下,多谢了!