IT之家 10 月 16 日消息,OpenAI 的 Codex / ChatGPT 负责人 @thsottiaux 昨日(10 月 15 日)在 X 平台发布推文,表示团队已上线一系列改动,并投入了更多算力,修复“所选模型已达容量上限,请尝试其他模型”报错。
IT之家注:OpenAI 团队当下正在推进“28 天持续交付”计划,在第 11 天公告中宣布已上线一系列改进并追加算力资源,目标是显著降低该错误的发生频率。
在难题反馈上,依据公开讨论看,该难题并非单纯的“模型真的满载”。有使用者报告称,即使订阅额度充足、切换不同模型,错误仍会反复浮现;也有开发者怀疑与后台请求、会话状态或资源调度有关。
在部署措施维度,OpenAI 团队借助优化平台,并投入更多算力,减少模型容量报错浮现,其次当报错仍然发生后,任务应能“最终自动继续”,而不是一直卡在错误状态、等待使用者干预。团队同时邀请仍在遇到难题的使用者反馈,以便进一步排查。
本文来源:https://www.ithome.com/1/013/753.htm
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。
IT House October 16 news, OpenAI's Codex/ChatGPT leader @ thsottiaux posted a tweet on X platform yesterday (October 15), saying that the team has launched a series of changes and invested more computing power to fix the error of "The selected model has reached the capacity limit, please try another model".
IT Home Note: The OpenAI team is currently moving forward with the "28 Days of Continuous Delivery" program, announcing in the 11th day announcement that a series of improvements and additional computing power resources have been launched, with the goal of significantly reducing the frequency of this error.

In terms of puzzle feedback, according to public discussion, the puzzle is not simply "the model is really full". Some users report repeated errors even when subscriptions are ample and different models are switched; others suspect background requests, session status, or resource scheduling.
In the dimension of deployment measures, the OpenAI team optimizes the platform and invests more computing power to reduce the error reporting of model capacity. Secondly, when the error reporting still occurs, the task should be able to "finally automatically continue", rather than being stuck in the error state and waiting for user intervention. The team also invites feedback from users who are still experiencing difficulties to further troubleshoot.
This article is from: https://www.ithome.com/1/013/753.htm For copyright issues, please write to: 13951431913#139.com or submit a review on the right.