deepinsight / deepinsight/insightface
when using other model, though loss fell normally, but lfw accuracy is 50%
- Dominant language
- Python
- Stars
- 29.7k
- Forks
- 6.1k
- PR merge metrics
- No merged PRs in 30d
Description
other model, eg., crossformer
https://github.com/cheerss/CrossFormer
INFO:root:Speed 480.88 samples/sec Loss 33.9955 LearningRate 0.0500 Epoch: 0 Global Step: 1900 Fp16 Grad Scale: 8192 Required: 69 hours
Training: 2021-09-23 09:56:18,663-Speed 480.88 samples/sec Loss 33.9955 LearningRate 0.0500 Epoch: 0 Global Step: 1900 Fp16 Grad Scale: 8192 Required: 69 hours
INFO:root:Speed 488.33 samples/sec Loss 33.4204 LearningRate 0.0500 Epoch: 0 Global Step: 1950 Fp16 Grad Scale: 8192 Required: 69 hours
Training: 2021-09-23 09:56:44,876-Speed 488.33 samples/sec Loss 33.4204 LearningRate 0.0500 Epoch: 0 Global Step: 1950 Fp16 Grad Scale: 8192 Required: 69 hours
INFO:root:Speed 468.55 samples/sec Loss 32.6157 LearningRate 0.0500 Epoch: 0 Global Step: 2000 Fp16 Grad Scale: 16384 Required: 69 hours
Training: 2021-09-23 09:57:12,194-Speed 468.55 samples/sec Loss 32.6157 LearningRate 0.0500 Epoch: 0 Global Step: 2000 Fp16 Grad Scale: 16384 Required: 69 hours
testing verification..
(12000, 512)
infer time 42.669530999999964
INFO:root:[lfw][2000]XNorm: 26.403959
Training: 2021-09-23 09:57:57,470-[lfw][2000]XNorm: 26.403959
INFO:root:[lfw][2000]Accuracy-Flip: 0.50000+-0.00000
Training: 2021-09-23 09:57:57,470-[lfw][2000]Accuracy-Flip: 0.50000+-0.00000
INFO:root:[lfw][2000]Accuracy-Highest: 0.50000
Training: 2021-09-23 09:57:57,470-[lfw][2000]Accuracy-Highest: 0.50000
Contributor guide
No contributing guide indexed for this repository
Assessment
This issue has not been assessed yet.