logical-reasoning / results /mgtv-llama3_p2_r3_full_metrics.csv
dh-mc's picture
ready for llama3 r4
fcc0071
raw
history blame
831 Bytes
epoch,model,accuracy,precision,recall,f1
0.0,hfl/llama-3-chinese-8b-instruct-v3_torch.bfloat16_lf,0.25066666666666665,0.6852419041932336,0.25066666666666665,0.32636449818329016
0.2,hfl/llama-3-chinese-8b-instruct-v3/checkpoint-35_torch.bfloat16_lf,0.7283333333333334,0.7722393813259697,0.7283333333333334,0.7426450360790026
0.4,hfl/llama-3-chinese-8b-instruct-v3/checkpoint-70_torch.bfloat16_lf,0.741,0.7868300593752113,0.741,0.7514058688729928
0.6,hfl/llama-3-chinese-8b-instruct-v3/checkpoint-105_torch.bfloat16_lf,0.6223333333333333,0.7771706776754249,0.6223333333333333,0.6762790454549326
0.8,hfl/llama-3-chinese-8b-instruct-v3/checkpoint-140_torch.bfloat16_lf,0.7,0.7767966010489314,0.7,0.7298480873851099
1.0,hfl/llama-3-chinese-8b-instruct-v3/checkpoint-175_torch.bfloat16_lf,0.697,0.78712001874989,0.697,0.7309586130328194