FanWan
FanWan
#include #include #include #include #include #include #include "aho_corasick.hpp" std::chrono::steady_clock::time_point time_start_point() { return std::chrono::steady_clock::now(); } double time_duration(const std::chrono::steady_clock::time_point& start) { auto end = time_start_point(); auto span = std::chrono::duration_cast(end - start).count(); return...
the outputs were as follows: construct time: 28.581 size of results: 2 match cost time: 53.299
Next, i did an experiment about exhaustive search using Python: `import os import time cur_dir = os.path.dirname(os.path.abspath(__file__)) + '/' data_dir = "/Users/wanfan01/workspace/cpp/learning/ac_machine/AhoCorasick/events.txt" if __name__ == "__main__": words_ = list() for...
crash occured in concurrent context, such as such as 100qps of reading threads and a crontab writing thread. Shared read lock is used when searching while exclusive write lock is...
> @wutingjun CUDA_VISIBLE_DEVICES=0,1 感觉这样去设置多卡是有存在问题的，我这样设置多卡的时候感觉比单卡的时间要多很多，请问这个问题，您解决了吗 请问一下，你们训练多卡微调的时候，有碰到学习率下降很快，才1.5个epoch学习率就为0了。
我有用Lora微调过，效果和稳定性都比chatglm1差。
> "code " in train data also generated by self-Instruct？ "code " in train data includes ''Instruction" and "Output" which can be constructed as Instruction-API pair ?
环境配置：cuda 11.4, Transformers 4.30.2, torch 2.0.1, Python 3.10
> 你好你有遇到过 ttributeError: 'ChatGLMModel' object has no attribute 'prefix_encoder' 报错吗 这个没有哈。
> py3.8/3.7 tf 4.27.1-4.29.2 torch1.13.1试试 requirements里面要求transformers==4.30.2，torch>=2.0