Model trained on training+validation set

[yankee624]

According to the readme, the released model is trained on Ego4D Narration Stream 113K + Ego4D GoalStep Stream 21K.
I guess it is trained with this script: scripts/ego4d/live1+.sh, which uses ego4d_goalstep_livechat_trainval, ego4d_refined_narration_stream_train, ego4d_refined_narration_stream_val, robustness to train the model.

But in Table1 of the paper, the model here seems to be trained only on ego4d_refined_narration_stream_train. Is this correct?
If so, can you please release this model?
The reason I'm requesting this is because: the currently released model is trained on ego4d_refined_narration_stream_train + ego4d_refined_narration_stream_val, so there is no remaining validation/test set I can use to test model performance. If there is other test set I can use, please let me know.