调用 azure上创建的text-embedding-ada-002模型，与调用原始gpt里的 text-embedding-ada-002。相同内容，生成结果不一样

Question

调用 azure上创建的text-embedding-ada-002模型，与调用原始gpt里的 text-embedding-ada-002。相同内容，生成结果不一样

hui zhong 0

我创建了text-embedding-ada-002这个模型，但是我使用相同内容，访问gpt原始api。获取到的embedding 向量，不一样。这个是什么原因呢

romungi-MSFT 48,916 Reputation points Microsoft Employee Moderator

2023-06-28T10:30:28.2833333+00:00

@hui zhong Adding the question translation in English.

I created the model text-embedding-ada-002, but I use the same content to access the original gpt api. The obtained embedding vector is different. What is the reason for this?

1 answer

Your answer

romungi-MSFT 48,916 Reputation points Microsoft Employee Moderator

2023-06-28T10:30:28.2833333+00:00

@hui zhong Adding the question translation in English.

I created the model text-embedding-ada-002, but I use the same content to access the original gpt api. The obtained embedding vector is different. What is the reason for this?

Answer 1

romungi-MSFT 48,916 Microsoft Employee Moderator

@hui zhong The difference in the responses from OpenAI and Azure OpenAI is primarily because of the training data of the models. In Azure OpenAI the training data of the models is mentioned in the model's page here. With text-embedding-ada-002 version 2 is recommended on Azure since it provides parity with OpenAI's text-embedding-ada-002. Please see the note from here.

If this answers your query, do click Accept Answer and Yes for was this answer helpful. And, if you have any further query do let us know.

Share via

调用 azure上创建的text-embedding-ada-002模型，与调用原始gpt里的 text-embedding-ada-002。相同内容，生成结果 不一样

1 answer