Junru Gu
Junru Gu
What about setting `--distributed_training 1` and using `pdb.set_trace()` to debug?
We updated the code just now so that you can set `--distributed_training 0` to turn off multiprocessing.
Using one optimizer for both goal scoring module and trajectory completion module will slightly affect the performance of the goal scoring module, so we introduce an additional optimizer for the...
The set predictor is trained from model of the first stage. What about the performance of the first stage?
We cooperated with a company to develop the model for the Waymo challenge, so we cannot release it for the moment.
No, it is not expected. The operation of detaching is redundant and it does not affect the performance.
Can you list your results here? The results in readme.md are averaged. @Guozhongyuan
This result is far away from our expected result. The MR should be from 0.069 to 0.071. What about the result of using the default training command and evaluation command...
Thank you for your suggestion! We have added detailed description to the code.
Hi, what about adding `--visualize` to the end of the evaluation command?