#7546 计算数据类型float16改成float32会不会更准.另外识别准确度数字调大有没有用

120.217**1 Posted at: 2 hours ago

计算数据类型float16改成float32会不会更准.另外识别准确度数字调大有没有用

1 Replies

#1 DeepSeek reply at 2 hours ago
  1. 计算数据类型改为 float32:会更准。

    float32 精度最高,但消耗资源多、速度慢;float16 适合 GPU,精度略低但速度快;int8 最省资源、速度最快,精度最低。

    如果显存充足且追求精度,可以改为 float32。

  2. 识别准确度(beam_size / best_of)数字调大:有提升作用。

    范围是 1-5,数字越大识别越准,但消耗显存越多、速度越慢;数字越小消耗显存越低。

    默认通常为 5(已较高),若显存吃紧可调低至 1。

Thinking...
Post Your Reply
If you have more than 50 messages, you will no longer receive an automatic reply. Please ask a new question!
Open source and free maintenance is not easy. If this project is helpful to you, please consider making a small donation to help the project continue to maintain and update.

Dondate via ko-fi.com

Related resource