CrisCloseTheDoor

> Hi @CrisCloseTheDoor, > > I've assigned this to another colleague of mine. In the meantime, can you share with us the full command that you ran? > > Ping...

I have solved this problem. The reason is that FP16 was used for high efficiency, but the overflow occurred in some addition operations, thus leaded to 'nan' data. A solution...

Thank u all, I 've tried @SwagJ and @wuhaoran111 's solution both and fix the same problem

I solved this probem by putting the main code into `def main()` and add the code in the end: `if __name__ == '__main__': main()`

> Hi, thanks for your reply. It is really helpful to have the checkpoints. I can see a similar performance as in your paper. > > I think the reason...

> Hi, thanks for your reply. It is really helpful to have the checkpoints. I can see a similar performance as in your paper. > > I think the reason...

By the way, the results in ETH/UCY are as good as paper's. Here're more details when running with deterministic JAAD: ``` Generating action sequence data fstride: 1 sample_type: all subset:...