基本流程
基于cpu的計算:
前往intel官網下載libmkl相關的庫:libmkl_core.a、libmkl_gf_lp64.a、libmkl_sequential.a
# 靜默安裝
./intel-onemkl-2025.2.0.629_offline.sh -a -s --eula accept
默認安裝目錄:/opt/intel/oneapi
libmkl相關的庫文件就在 /opt/intel/oneapi/mkl/latest/lib/
數據
http://hltc.cs.ust.hk/iwslt/index.php/evaluation-campaign/ted-task.html
模型
代碼
標題:A Context-Aware Incremental Model for Punctuation Prediction: Industrial Deployment in Large-Scale Concurrent and Long-Text Simultaneous Translation Scenarios
這句話,當前好多大模型都翻譯錯誤:I want a flight to Boston um to Denver
包括豆包
還有deepseek:
但是gemini的表現更好: