LLM_VISION_IMAGE_USE_BASE64=1 是否会导致tokens的大量消耗?
#6731
SAnBlog
started this conversation in
General | 讨论
Replies: 1 comment 3 replies
-
|
上传图片的会走专门的图片token计算逻辑，和base64上传/url上传没有区别 |
Beta Was this translation helpful? Give feedback.
3 replies
-
那url上传的优势是什么? |
Beta Was this translation helpful? Give feedback.
-
|
base64 上传会有接口请求流量的开销。比如你传一张 10M 的 base64 图片到服务商，如果是后端传输，那么就会产生 10m 的出口流量。如果你的出口流量有带宽限制，比如最多 1m/s ，那么你就会需要等待 10s 才能传完毕。如果是纯 url 的话，那么图片下载逻辑是会在服务商侧完成，一般提供服务的也都是大带宽，速度会比较快。 |
Beta Was this translation helpful? Give feedback.
-
|
明白了感谢解答,对于带宽足够自用的影响不大. |
Beta Was this translation helpful? Give feedback.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
目前Gemini使用OneApi中专无法上传图片,看到可以采用LLM_VISION_IMAGE_USE_BASE64的解决方案,确实解决了问题,但是有个疑问,既然是将图片转成base64,图片的大小转base64后的长度接近,那岂不是2M的一张图片会消耗大量的tokens?
有些模型上下文长度不足1M的能否上传2M的大小?
Beta Was this translation helpful? Give feedback.
All reactions