How to interpet evaluation results?

[levTepl]

Hello,

I've trained a dummy model for 30 epochs and want to evaluate it.
As a result I get the following output:

Generate Logit: [130/157]
Generate Logit: [140/157]
Generate Logit: [150/157]
Logits calculation time: 1.439882755279541
Val acc      Test acc
0.465       0.516
0.537       0.566
0.583       0.605
0.601       0.631

*********************
valid acc: 46.880, test acc: 51.780, test top5 acc: 80.160 nlpd: 1.841, ECE: 0.059, test flops: 7.03
*********************
valid acc: 47.620, test acc: 52.300, test top5 acc: 80.610 nlpd: 1.827, ECE: 0.061, test flops: 7.29
*********************
valid acc: 48.400, test acc: 52.650, test top5 acc: 81.050 nlpd: 1.815, ECE: 0.065, test flops: 7.60
*********************
valid acc: 49.240, test acc: 53.050, test top5 acc: 81.430 nlpd: 1.803, ECE: 0.068, test flops: 7.96

My question is - how to interpret this output? Too many different accuracies :)
I'd like to reproduce these results.