eval: retry in more scenarios
This commit is contained in:
parent
47af878ebb
commit
a2ce038352
1 changed files with 59 additions and 38 deletions
|
@ -1663,13 +1663,19 @@ async fn retry_on_rate_limit<R>(mut request: impl AsyncFnMut() -> Result<R>) ->
|
|||
attempt += 1;
|
||||
match request().await {
|
||||
Ok(result) => return Ok(result),
|
||||
Err(err) => match err.downcast::<LanguageModelCompletionError>() {
|
||||
Err(err) => {
|
||||
if attempt > 20 {
|
||||
return Err(err);
|
||||
}
|
||||
|
||||
match err.downcast::<LanguageModelCompletionError>() {
|
||||
Ok(err) => match &err {
|
||||
LanguageModelCompletionError::RateLimitExceeded { retry_after, .. }
|
||||
| LanguageModelCompletionError::ServerOverloaded { retry_after, .. } => {
|
||||
let retry_after = retry_after.unwrap_or(Duration::from_secs(5));
|
||||
// Wait for the duration supplied, with some jitter to avoid all requests being made at the same time.
|
||||
let jitter = retry_after.mul_f64(rand::thread_rng().gen_range(0.0..1.0));
|
||||
let jitter =
|
||||
retry_after.mul_f64(rand::thread_rng().gen_range(0.0..1.0));
|
||||
eprintln!(
|
||||
"Attempt #{attempt}: {err}. Retry after {retry_after:?} + jitter of {jitter:?}"
|
||||
);
|
||||
|
@ -1693,7 +1699,21 @@ async fn retry_on_rate_limit<R>(mut request: impl AsyncFnMut() -> Result<R>) ->
|
|||
|
||||
// Use server-provided retry_after if available, otherwise use default
|
||||
let retry_after = retry_after.unwrap_or(Duration::from_secs(5));
|
||||
let jitter = retry_after.mul_f64(rand::thread_rng().gen_range(0.0..1.0));
|
||||
let jitter =
|
||||
retry_after.mul_f64(rand::thread_rng().gen_range(0.0..1.0));
|
||||
eprintln!(
|
||||
"Attempt #{attempt}: {err}. Retry after {retry_after:?} + jitter of {jitter:?}"
|
||||
);
|
||||
Timer::after(retry_after + jitter).await;
|
||||
continue;
|
||||
}
|
||||
LanguageModelCompletionError::ApiInternalServerError { .. }
|
||||
| LanguageModelCompletionError::ApiReadResponseError { .. }
|
||||
| LanguageModelCompletionError::DeserializeResponse { .. }
|
||||
| LanguageModelCompletionError::HttpSend { .. } => {
|
||||
let retry_after = Duration::from_secs(attempt);
|
||||
let jitter =
|
||||
retry_after.mul_f64(rand::thread_rng().gen_range(0.0..1.0));
|
||||
eprintln!(
|
||||
"Attempt #{attempt}: {err}. Retry after {retry_after:?} + jitter of {jitter:?}"
|
||||
);
|
||||
|
@ -1703,7 +1723,8 @@ async fn retry_on_rate_limit<R>(mut request: impl AsyncFnMut() -> Result<R>) ->
|
|||
_ => return Err(err.into()),
|
||||
},
|
||||
Err(err) => return Err(err),
|
||||
},
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
|
Loading…
Add table
Add a link
Reference in a new issue