Is the embedding service overloaded?

API
1

I’m using text-embedding-ada-002.
this error always happen after several calls.

    772     return self._value
    773 else:
--> 774     raise self._value

ServiceUnavailableError: The server is overloaded or not ready yet.
2

Could you share the rest of you code as well, including the function call you are making to ada-002 model ?

3

There was another report of issues seven hours before Connection to: OpenAI API failed with status: 503 error: Service Unavailable

Ran a batch of calls again just to test: is fine.

PS, running identical embeddings as a list in-context on a single call, what a non-determinism looks like at about 15% of runs (samples are middle dimensions):

[Total tokens for 2 embeddings] Counted 6; API said 6
== Sample from vectors returned ==
0 [-0.04312577471137047, 0.011266510933637619, -0.01897585764527321]
1 [-0.043195661157369614, 0.01125156506896019, -0.01903713122010231]
== Cosine similarity and vector comparison of all inputs ==
0:‘cute kitten’ <==> 1:‘cute kitten’:
0.9999972582 0.0000000000- identical: False

== Sample from vectors returned ==
0 [-0.043184179812669754, 0.01122297067195177, -0.018991723656654358]
1 [-0.04312577471137047, 0.011266510933637619, -0.01897585764527321]
== Cosine similarity and vector comparison of all inputs ==
0:‘cute kitten’ <==> 1:‘cute kitten’:
0.9999973774 0.0000000000- identical: False

1 Like
5

This error happens again today when I’m trying to use text-embedding-ada-002.
my code :

def embedd_it(text):
    apikey="XXX"
    openai.api_key=apikey
    return openai.Embedding.create(model="text-embedding-ada-002",input=text)['data'][0]['embedding']


df_train['embeddings'] = df_train['content'].parallel_apply(embedd_it)

same error message:


ServiceUnavailableError: The server is overloaded or not ready yet.

my openai version: 0.27.7

6

How big is your training data ? When you’re using implementing parellel reply, there is a chance it might be crossing you Rate limit, which might be causing the server overload

7

I have been getting the same error the past couple of days and i am not even close to the rate limit

9

try upgrade your openai version to the newest. And change your method to client.XXX as the new document does. I solved my problem by doing this.