stage2_train.log

========== 12151654 - train_reranker_v78.yaml ==========

可用的 GPU 数量: 1
polars==1.12.0
torch==2.5.1+cu124
transformers==4.42.4
datasets==3.1.0
sentence_transformers==3.2.1

train instruction_token_len range: 169 ~ 533
valid instruction_token_len range: 176 ~ 547

start training...
finish training...
Training loss: 0.26723508124220563