The log does not contain the evaluated value. Can the epoch be made smaller? Which indicator represents Overall performance?

[mrsempress]

1. The log file only contains loss information, and the evaluation information is only printed in the terminal, which is very unfriendly for epoch 400 training. I have to test each epoch again.
2. Is epoch 400 necessary? In my experiments, val loss became overfitting near epoch 50. Can the epoch be adjusted to 75 or other relatively small values? Will it affect the results? Have you done any corresponding experiments?
3. How long does it take to run the complete 400 epochs?
4. In your public training log, it only reaches epoch 72. In issue 3, I find you load epoch 60, and the overall result is from last_semantic alignment Acc0.25: Top-1: 0.54586. But in the training log, the results of epoch 60 are:

last_ Box given span (soft-token) Acc0.50: Top-1: 0.000, Top-5: 0.000, Top-10: 0.000
last_ Box given span (contrastive) Acc0.25: Top-1: 0.546, Top-5: 0.680, Top-10: 0.736
last_ Box given span (contrastive) Acc0.50: Top-1: 0.423, Top-5: 0.573, Top-10: 0.627
proposal_ Box given span (soft-token) Acc0.25: Top-1: 0.000, Top-5: 0.000, Top-10: 0.000
proposal_ Box given span (soft-token) Acc0.50: Top-1: 0.000, Top-5: 0.000, Top-10: 0.000
proposal_ Box given span (contrastive) Acc0.25: Top-1: 0.000, Top-5: 0.000, Top-10: 0.000
proposal_ Box given span (contrastive) Acc0.50: Top-1: 0.000, Top-5: 0.000, Top-10: 0.000
0head_ Box given span (soft-token) Acc0.25: Top-1: 0.000, Top-5: 0.000, Top-10: 0.000
0head_ Box given span (soft-token) Acc0.50: Top-1: 0.000, Top-5: 0.000, Top-10: 0.000
0head_ Box given span (contrastive) Acc0.25: Top-1: 0.000, Top-5: 0.000, Top-10: 0.000
0head_ Box given span (contrastive) Acc0.50: Top-1: 0.000, Top-5: 0.000, Top-10: 0.000
1head_ Box given span (soft-token) Acc0.25: Top-1: 0.000, Top-5: 0.000, Top-10: 0.000
1head_ Box given span (soft-token) Acc0.50: Top-1: 0.000, Top-5: 0.000, Top-10: 0.000
1head_ Box given span (contrastive) Acc0.25: Top-1: 0.000, Top-5: 0.000, Top-10: 0.000
1head_ Box given span (contrastive) Acc0.50: Top-1: 0.000, Top-5: 0.000, Top-10: 0.000
2head_ Box given span (soft-token) Acc0.25: Top-1: 0.000, Top-5: 0.000, Top-10: 0.000
2head_ Box given span (soft-token) Acc0.50: Top-1: 0.000, Top-5: 0.000, Top-10: 0.000
2head_ Box given span (contrastive) Acc0.25: Top-1: 0.000, Top-5: 0.000, Top-10: 0.000
2head_ Box given span (contrastive) Acc0.50: Top-1: 0.000, Top-5: 0.000, Top-10: 0.000
3head_ Box given span (soft-token) Acc0.25: Top-1: 0.000, Top-5: 0.000, Top-10: 0.000
3head_ Box given span (soft-token) Acc0.50: Top-1: 0.000, Top-5: 0.000, Top-10: 0.000
3head_ Box given span (contrastive) Acc0.25: Top-1: 0.000, Top-5: 0.000, Top-10: 0.000
3head_ Box given span (contrastive) Acc0.50: Top-1: 0.000, Top-5: 0.000, Top-10: 0.000
4head_ Box given span (soft-token) Acc0.25: Top-1: 0.000, Top-5: 0.000, Top-10: 0.000
4head_ Box given span (soft-token) Acc0.50: Top-1: 0.000, Top-5: 0.000, Top-10: 0.000
4head_ Box given span (contrastive) Acc0.25: Top-1: 0.000, Top-5: 0.000, Top-10: 0.000
4head_ Box given span (contrastive) Acc0.50: Top-1: 0.000, Top-5: 0.000, Top-10: 0.000

Why is one segment and one box?

[mrsempress]

5. In addition, I would like to know the specific meanings of last_ Box and proposal_ Box, soft-token and contrastive.
6. I want to know which values should correspond to the [email protected](det) and [email protected](det) results of Unique (∼19%) and Multiple (∼81%).

[mrsempress]

For question 6, is it from Analysis Unique and multi?

[mrsempress]

7. it reports last_ position and last_ semantic in validation. What are their respective meanings?

[mrsempress]

8. Under what circumstances is it appropriate to stop training, or in other words, the optimal performance is generally around the epoch number

[mrsempress]

9. How long is the testing time? It takes me 45 minutes here. But in your log is 11'20. I use the command srun -p test --gres=gpu:1 sh test_scanrefer.sh. What did I do wrong?