内测期间:充值额外赠送 30% credits,订阅额外赠送 50%–70%。有任何意见或建议,欢迎到支持页反馈。 Closed beta: top-ups come with 30% extra credits, subscriptions with 50–70% extra. We welcome your feedback on the support page.

提交任务Submit a job

01代码来源Source code

把整个目录打包成 .zip,私有代码请用这种方式。脚本中的入口文件写包内相对路径。大小上限 64 MB;数据请勿打进代码包,改用下方的数据集上传。 Zip the entire directory — use this option for private code. In the script, reference the entry file by its path inside the archive. 64 MB max; upload data through the dataset field below rather than bundling it here.

支持 .zip / .tar / .tar.gz,上限 1 GB。运行前会解压到该路径,脚本照常引用即可(例如 --data ./data)。数据只进沙箱,不会出现在调度机、报告或补丁中,7 天后自动删除。 .zip, .tar or .tar.gz, up to 1 GB. It is extracted to this path before the run, so your script references it as usual (e.g. --data ./data). The data stays inside the sandbox — it never reaches the dispatcher, the report or the patch — and is deleted automatically after 7 days.

02运行方式How to run it

点击行号指定测速行,再次点击恢复自动识别Click a line number to set the measured line; click again to revert to automatic

和在自己机器上一样写:安装依赖、下载数据集(wget / curl / huggingface-cli)、设置环境变量,最后一行写你平时运行的命令。 加速比只针对测速行做端到端复测(默认取最后一条以 python / torchrun / accelerate … 开头的命令); 测速行之前的部分视为环境准备:只执行一次,不计时,执行时不分配显卡。 Write it exactly as you would on your own machine: install dependencies, fetch the data (wget / curl / huggingface-cli), set environment variables, and finish with the command you normally run. The speedup is measured end to end on the measured line only (by default, the last command starting with python / torchrun / accelerate …); everything before it is treated as setup — run once, untimed, and without a GPU.

高级选项Advanced options

选 auto 会在 RTX 4090 上运行,便宜,也能满足大多数模型;显存不够时再在这里换更大的卡。 加速比基于同一张卡上的基线计算,换大卡不会让数字虚高,只会更贵。 如果程序本身不用显卡(CSV 处理、解码、纯 numpy 流水线),请选「CPU only」:不收机器费,只按 token 计费。 Auto runs on an RTX 4090, which is inexpensive and fits most models. Choose a larger GPU here if your model does not fit. If your program never uses a GPU (a CSV tool, a decoder, a pure-NumPy pipeline), choose "CPU only": no machine fee, and tokens are priced as usual. The speedup is measured against that same GPU's own baseline, so a larger GPU does not inflate the number — it only costs more.

同一个仓库可以分别用两个模型各跑一次做对比,token 计价方式相同。 You can run the same repository once with each agent to compare them; tokens are priced identically.

03开始方式Get started

几秒钟的静态扫描,不扣费,也不执行你的代码。看到预估加速比后,再决定是否继续。 A static scan that finishes in seconds. Nothing is charged and none of your code runs. Review the projected speedup, then decide whether to continue.

先免费估算(几秒钟,不执行代码)。点「继续」时扣 2 credits 分析押金;没有加速时退回,失败在我们这边时同样退回(基础设施故障、超时、重新部署中断、环境搭建失败、找不到入口实际运行的代码、程序显存超出我们最大的显卡)。解锁补丁时全额抵扣,运行按实际用量计费。 A free estimate first (seconds, none of your code runs). Continuing takes the 2-credit analysis deposit, returned when no speedup is found or when the failure is ours: infrastructure failure, a timeout, a redeploy interrupting the run, the environment failing to build, an entry point that dispatches to code we cannot find, or a program that needs more GPU memory than our biggest GPU — and credited in full against the patch; the run bills on actual usage.

账户Account

用量包 credits 自购买起 24 个月内有效;订阅 credits 在每个周期结束时清零。 Usage-pack credits are valid for 24 months from purchase; subscription credits reset at the end of each period.

充值与订阅Add credits & plans

读取中…Loading…

最近账目Recent activity

读取中…Loading…

邀请Invite

读取中…Loading…

邮件通知Email notifications

读取中…Loading…

完整账单明细 →Full statement → · 全部方案与费率All plans & rates