pcam-interpretability/dino-vits16-val08398-vit-tuned-safe

dino-vits16

Best Validation Accuracy: 0.8398

Metadata

• Model Name: dino-vits16
• Optimizer: adamw
• Scheduler: cosine
• Weight Decay: 0.001
• Warmup Epochs: 5
• Patience: 10
• Amp: True
• Seed: 42
• Batch Size: 352
• Initial Lr: 0.001
• Total Epochs Ran: 38
• Early Stopped: True
• Training Time Seconds: 22297.275145292282
• Num Parameters: 21666049
• Device: NVIDIA A100-SXM4-40GB
• Run Id: vit-tuned-safe

Training Configuration

• Epochs: 38
• Batch size: 352
• Learning rate (initial): 0.001

Training Logs (Per Epoch)

Epoch	Train Loss	Train Acc	Val Loss	Val Acc	LR
1	0.6211	0.7119	0.4845	0.7484	0.000200
2	0.4682	0.7774	0.5087	0.7466	0.000400
3	0.4455	0.7918	0.4857	0.7465	0.000600
4	0.4391	0.7945	0.4634	0.7683	0.000800
5	0.4309	0.7989	0.4192	0.7904	0.001000
6	0.4169	0.8078	0.4200	0.7889	0.001000
7	0.3996	0.8167	0.3879	0.8149	0.000999
8	0.3911	0.8222	0.4468	0.7795	0.000996
9	0.3830	0.8264	0.3944	0.8103	0.000991
10	0.3759	0.8305	0.4086	0.7977	0.000984
11	0.3678	0.8353	0.4321	0.7857	0.000976
12	0.3621	0.8377	0.4163	0.8022	0.000965
13	0.3574	0.8411	0.3871	0.8170	0.000952
14	0.3543	0.8418	0.5018	0.7661	0.000938
15	0.3490	0.8453	0.4141	0.8099	0.000922
16	0.3412	0.8499	0.3623	0.8295	0.000905
17	0.3336	0.8534	0.4005	0.8193	0.000885
18	0.3565	0.8418	0.3622	0.8311	0.000864
19	0.3494	0.8447	0.3668	0.8304	0.000842
20	0.3359	0.8521	0.4189	0.8000	0.000819
21	0.3355	0.8519	0.3609	0.8314	0.000794
22	0.3280	0.8566	0.3757	0.8241	0.000768
23	0.3203	0.8606	0.3917	0.8174	0.000741
24	0.3278	0.8571	0.3974	0.8180	0.000713
25	0.3109	0.8649	0.3669	0.8310	0.000684
26	0.3129	0.8649	0.3511	0.8390	0.000655
27	0.3080	0.8667	0.3574	0.8373	0.000624
28	0.3050	0.8686	0.3584	0.8398	0.000594
29	nan	0.8688	nan	0.8383	0.000563
30	nan	0.6657	nan	0.5005	0.000531
31	nan	0.5000	nan	0.5005	0.000500
32	nan	0.5000	nan	0.5005	0.000469
33	nan	0.5000	nan	0.5005	0.000437
34	nan	0.5000	nan	0.5005	0.000406
35	nan	0.5000	nan	0.5005	0.000376
36	nan	0.5000	nan	0.5005	0.000345
37	nan	0.5000	nan	0.5005	0.000316
38	nan	0.5000	nan	0.5005	0.000287