Fixing bug with max tokens
This commit is contained in:
parent
fbca788cbf
commit
588b0a434d
@ -165,10 +165,16 @@ class Pipeline:
|
|||||||
}
|
}
|
||||||
else:
|
else:
|
||||||
reasoning_config = {}
|
reasoning_config = {}
|
||||||
|
|
||||||
|
# If budget_tokens is greater than max_tokens, adjust max_tokens to MAX_COMBINED_TOKENS
|
||||||
|
max_tokens = body.get("max_tokens", MAX_COMBINED_TOKENS)
|
||||||
|
if max_tokens < budget_tokens and budget_tokens > 0:
|
||||||
|
max_tokens = MAX_COMBINED_TOKENS
|
||||||
|
|
||||||
payload = {"modelId": model_id,
|
payload = {"modelId": model_id,
|
||||||
"messages": processed_messages,
|
"messages": processed_messages,
|
||||||
"system": [{'text': system_message['content'] if system_message else 'you are an intelligent ai assistant'}],
|
"system": [{'text': system_message['content'] if system_message else 'you are an intelligent ai assistant'}],
|
||||||
"inferenceConfig": {"temperature": 1, "maxTokens": body.get("max_tokens", MAX_COMBINED_TOKENS)},
|
"inferenceConfig": {"temperature": 1, "maxTokens": max_tokens },
|
||||||
"additionalModelRequestFields": reasoning_config
|
"additionalModelRequestFields": reasoning_config
|
||||||
}
|
}
|
||||||
if body.get("stream", False):
|
if body.get("stream", False):
|
||||||
|
Loading…
x
Reference in New Issue
Block a user