statsu1990
commented
4 years ago (epoch, warmup epoch, learning rate, different learning rate)
- Model_v1_2_1, (3,0,3e-5,x1) : score 0.547774 (only posi and nega)
- Model_v1_3_0, (3,0,3e-5,x100) : score 0.546235 (only posi and nega)
- Model_v1_3_1, (3,1,3e-5,x100) : score 0.540208 (only posi and nega)
- Model_v1_3_2, (3,0,3e-5,x30) : score 0.550897 (only posi and nega)
- Model_v1_3_3, (3,0,1e-5,x100) : score 0.547941 (only posi and nega)
- Model_v1_3_4, (3,0,1e-5,x30) : score 0.553748, 0.551891 (only posi and nega)
- Model_v1_3_5, (3,1,1e-5,x100) : score 0.544455 (only posi and nega)
- Model_v1_3_6, (4,1,1e-5,x30) : score 0.547059 (only posi and nega)
https://www.kaggle.com/c/google-quest-challenge/discussion/129840 https://qiita.com/YuiKasuga/items/343309257da1798c1b63