题意:“Azure OpenAI 模型中的 GPT-3.5-Turbo 每分钟的请求限制是多少?”
问题背景:
I couldn't find the RPM limit for Azure openAI
“我找不到 Azure OpenAI 的每分钟请求限制。”
I found the TPM limit thou which is 240k for East US, South Central US, West Europe, France Central, UK South,
“不过我找到了每分钟请求数的限制,即东美、美国南中部、西欧、法国中部和英国南部的限制是 240k。”
I've tried Microsoft documentation but couldn't find this limit. Azure OpenAI Service quotas and limits - Azure AI services | Microsoft Learn
“我查阅了微软文档,但没能找到这个限制。Azure OpenAI 服务配额和限制 - Azure AI 服务 | Microsoft Learn”
问题解决:
The "RPM" (request per minute) quota is directly a rate of the "TPM" (transaction per minute):
“‘RPM’(每分钟请求数)配额直接是‘TPM’(每分钟事务数)的速率。”
So for GPT 3.5 Turbo models, if you have a 240k TPM for your deployment, the RPM will be 1440 requests per minute.
“因此,对于 GPT-3.5 Turbo 模型,如果您的部署有 240k TPM,那么 RPM 将是每分钟 1440 次请求。”
Example visible in Azure AI Studio: in my case, I set a 120k TPM, I got 720 RPM:
“在 Azure AI Studio 中可以看到的示例:以我的情况为例,我设置了 120k TPM,得到了 720 RPM。”
Source: Manage Azure OpenAI Service quota - Azure AI services | Microsoft Learn