Rangehow

I think there is a misunderstanding regarding the accuracy aspect of what I raised. What I am comparing is the tensor results generated by encoding the same text input using...

I assume code should be like below in `client.py` ```python http_client = httpx.AsyncClient( limits=httpx.Limits(max_connections=2048, max_keepalive_connections=1024), timeout=httpx.Timeout(timeout=None) ) @limit_async_func_call(max_size=1024) # 限制并发为1024 async def custom_httpx(prompt, system_prompt=None, history_messages=[], **kwargs): messages = [] if...

I just tested two from scratch OpenAI Async API:  HTTPX: （I got distracted and didn’t notice it had been running for quite a while.） 

TypeError: ModernBertModel.__init__() got an unexpected keyword argument 'reference_compile'

> 2. Can we use [Structured Output](https://platform.openai.com/docs/guides/structured-outputs/introduction) from OpenAI to ensure the response follows are data models and we don't have to retry multiple times. (saves cost and latency) >...