def get_chat_response()

in lmms_eval/tasks/mathvista/mathvista_evals.py [0:0]

• 31 lines of code
• 13 McCabe index (conditional complexity)

    def get_chat_response(self, prompt, temperature=0, max_tokens=256, n=1, patience=10000000, sleep_time=0):
        messages = [
            {"role": "user", "content": prompt},
        ]
        payload = {"model": self.gpt_model, "messages": messages, "temperature": temperature, "max_tokens": max_tokens, "n": n}

        while patience > 0:
            patience -= 1
            try:
                response = self._post_request(payload)
                if n == 1:
                    prediction = response["choices"][0]["message"]["content"].strip()
                    if prediction and prediction != "":
                        return prediction
                else:
                    prediction = [choice["message"]["content"].strip() for choice in response["choices"]]
                    if prediction and prediction[0] != "":
                        return prediction

            except Exception as e:
                if "Rate limit" not in str(e):
                    eval_logger.error(e)

                if "Please reduce the length of the messages" in str(e):
                    eval_logger.error("!!Reduce prompt size")
                    # reduce input prompt and keep the tail
                    new_size = int(len(prompt) * 0.9)
                    new_start = len(prompt) - new_size
                    prompt = prompt[new_start:]
                    payload["messages"] = [
                        {"role": "user", "content": prompt},
                    ]

                if sleep_time > 0:
                    time.sleep(sleep_time)
        return ""