DarrenWang
DarrenWang
@PkuRainBow Hi, could you share your training configuration, such as learning rate, data augmentation?
后面怎么样了呢？我也是移植到arm上就错了，在pc上都是好的。
定位到了问题，是mat里的数据归一化有bug
float32x4_t _mean = vdupq_n_f32(mean); for (; nn>0; nn--) { float32x4_t _ptr = vld1q_f32(ptr); _ptr = vsubq_f32(_ptr, _mean); vst1q_f32(ptr, _ptr); ptr += 4; } 这里的_ptr需要除以std.
> hi, I had a problem at converting my 3D semantic segmentation result to the submission format. May I know how you did it? Thank you so much ! You...
> Hi @jinfagang, > Sorry for the late reply. > This is weird. Did you set the batch size as 1? > The information in README is out of date....
> > 请问是做了什么改动吗？ > > 14B的话，模型参数是没有更新的，相关代码上有一些模型间统一的变化，主要是速度相关的优化。这部分变动应该不会影响训练结果才对，方便说下loss大概高了多少吗？ 之前训练完的loss大概是0.001x，现在试了几次是0.03x