ZikangZhou / QCNet

[CVPR 2023] Query-Centric Trajectory Prediction
https://openaccess.thecvf.com/content/CVPR2023/papers/Zhou_Query-Centric_Trajectory_Prediction_CVPR_2023_paper.pdf
Apache License 2.0
481 stars 76 forks source link

Is this what the training should look like ? #6

Closed shahaamirbader closed 1 year ago

shahaamirbader commented 1 year ago

Hi I have been training the model for two days now on 2 x V100 GPUs 32 GB. This is what the output at my terminal looks like. Can you comment on if the training is going in the right direction? I see some losses increasing here. Also how many epochs is the training suppose to last for?

image

ZikangZhou commented 1 year ago

The training looks good. In your screenshot, isn't the loss generally decreasing (some losses are negative)? You can monitor the training process using tensorboard or wandb.

The default max_epoch is set to 64, but you can adjust those training hyperparameters, as it seems that your batch size is currently set to 8.

HIT-five commented 1 year ago

How long will the model training last if I use the 8 NVIDIA 3090? thank you~

ZikangZhou commented 1 year ago

How long will the model training last if I use the 8 NVIDIA 3090? thank you~

Two days I think.