tp submit

tp submit <EXP> [options]

本地校验并上传实验，展示服务端解析的预览，确认后才创建排队任务。无人值守提交使用 --yes。完整流程见提交作业。这一页是参数参考，下面那张表来自 CLI 本身，所以它不可能过期。

实例

# 8 张卡上的 NeMo-RL GRPO
tp submit my-grpo --profile h200:8

# verl，配平台数据集和一次性的超参覆盖
tp submit my-verl --profile h200:4 \
  --model Qwen/Qwen3.5-9B \
  --train-dataset alice/gsm8k-zh@v2 --train-data train.parquet \
  -s actor_rollout_ref.actor.optim.lr=1e-6

# 异构多 pool，训练成功后导出
tp submit my-grpo --profile train=h200:8 --profile rollout=h100:2 --then export

# 从前一次运行的 checkpoint 继续
tp submit my-dpo --profile h200:8 --init-from run/run-4f2a91/checkpoint

# 你自己的训练器
tp submit my-custom --profile h100:1 \
  --image registry.example.com/train@sha256:abcd… \
  --observability-url https://wandb.example/my-proj

参数表装不下的几点

项目名来自 tuneplane.yaml 里的 name。没有 --project。
--framework-version 只有和 --upgrade-recipe 一起用才有效。
声明 external observability 的 recipe 必须带 --observability-url，声明 platform 的会拒绝它。
--volume 不带版本后缀，--environment 必须带。
给了 --train-dataset 之后，--train-data 是数据集内部的路径。

保存预览与重试

tp submit my-grpo --profile h200:8 --preview
# 检查命令输出的 JSON 文件后再确认：
tp submit --commit .tuneplane-data/submissions/<preview-id>.json
# 无人值守提交：
tp submit my-grpo --profile h200:8 --yes

预览有效期为一小时。确认使用预览 ID、摘要和固定重试键。响应丢失时，请用同一个文件重复执行 --commit；已成功提交的回执不受预览过期影响。重新生成预览代表一次新提交，可能创建另一个任务。 --intent intent.json 支持三种输入，每种都需要 form、project 和完整的 tuneplane/v1 JobSpec（spec）：

输入方式	补充字段
`source`	`source_file`：相对于意图文件的本地 `.tar.gz` 路径
`recipe`	`config`：JSON 对象形式的完整原生配置
`image`	`argv`：命令参数数组；可选 `working_directory`，默认 `/workspace`

镜像方式使用 custom/custom 配方。参数按字面值传递；如需 shell，必须显式指定。 OCI 镜像必须在 JobSpec 或部署侧运行时注册表中使用 @sha256: 固定内容。部署侧需要提供独立 runner 文件。凭据使用平台管理的绑定，不要写入命令参数或配置文件。任务页面提供相同的三种输入、解析预览、独立确认、预览保存和代码包下载。浏览器源码上传上限为 256 MiB，更大的归档使用 CLI；CLI 遵循部署配置的上传上限。预览和下载都不会启动任务。 API 依次使用 POST /api/submissions/uploads、带身份认证和 X-Upload-Token 的 PUT /api/submissions/uploads/{id}、POST /api/submissions/previews 和 POST /api/submissions/commits。GET /api/submissions/previews/{id}/package 导出本人未过期预览的代码包。确认后，任务所有者还可通过 GET /api/jobs/{run_id}/submission/package 导出，不受预览过期影响。旧的归档／请求头 POST /api/jobs 协议已停用，返回 410。

`tp submit`

Submit a training job

tp submit [EXP] [options]

参数	说明
`EXP`	Experiment name or path

选项	说明
`--profile`	Target hardware and resources in one argument: h200 (the registry default shape), h200:4 (4 GPUs), h200:16 (2 full nodes). Repeat it to split pools per role (the heterogeneous extension point): —profile train=h200:8 —profile rollout=h100:2. Comma-separate roles that share a pool: —profile actor,rollout=h200:8 —profile teacher=h200:8 (可重复)
`--method`, `-m`	Method identifier <framework>/<method> (e.g. nemo-rl/grpo, verl/grpo); read from the experiment directory’s recipe.lock.json when omitted
`--set`, `-s`	Override a hyperparameter, repeatable. Types and ranges are checked locally against the method declaration, so a typo fails immediately (可重复)
`--init-from`	Start from the previous stage’s artifact (the SFT -> DPO -> GRPO pipeline)
`--model`	Base model path or Hub id; required for verl/TRL
`--train-data`	Training data path; required for verl/TRL. With a platform dataset declared (in the config or via —train-dataset), write the file name relative to the dataset (e.g. train.parquet) and the job side resolves it into the cache directory
`--validation-data`	Validation data path; required for verl/TRL, used like —train-data
`--train-dataset`	Platform dataset reference: pulled into the shared cache when the job starts, with <NAME>_DATA_DIR injected. Prefer declaring it as data.train.dataset in the experiment config; this flag is only a temporary override
`--validation-dataset`	Platform dataset reference for the validation set; matches data.validation.dataset in the config
`--output-dataset`	Publish what this run produces as this dataset version. The job writes files into TUNEPLANE_DATASET_OUT_DIR and the platform publishes them when the run succeeds — the job is never given a credential for the registry, and the published dataset inherits the classification of the material the run read. Used by a preprocessing run over a corpus
`--volume`	Mount a governed directory read-only at $VOLUMES_DIR/<name>; repeatable. Prefer declaring it as data.volumes in the experiment config; this flag is only a temporary override. A volume has no version — its current contents are its contents. Run `tp volume ls` for the ids (可重复)
`--environment`	Agent RL environment reference. Delivered like a corpus and for a stronger reason: a private environment is non-exportable, so the platform resolves the version into the job runtime and mounts it read-only. Run `tp env ls` to see the environments you can reference
`--taskset`	Immutable taskset <owner>/<name>@<version>
`--image`	Job image (a tag is enough, e.g. registry/nemo-rl-ext:v0.7.0). Required for custom; for other frameworks it overrides the default image the console configured for them
`--then`	Run automatically after training succeeds (export/eval), repeatable (可重复)
`--observability-url`	For an external observability recipe only; a platform recipe must not set it
`--framework-version`	Only together with —upgrade-recipe; on its own it does not rewrite the lock file
`--upgrade-recipe`	Upgrade this experiment’s lock to the current catalog before submitting, reusing tp recipe upgrade
`--allow-dirty`	Allow uncommitted changes in the working tree (refused by default, to keep submissions traceable)
`--no-validate`	Skip the pre-submission validation
`--intent` `<file>`	Recipe, source or image submission JSON; source_file is resolved relative to this file
`--commit` `<file>`	Confirm or retry a saved preview without uploading or resolving again
`--preview`	Resolve and save a preview without creating a Job
`--yes`, `-y`	Confirm the resolved submission without prompting

CLI 参考

配置

实验

训练

产物与评测

资源

管理

实例

参数表装不下的几点

保存预览与重试

`tp submit`

​实例

​参数表装不下的几点

​保存预览与重试

​tp submit

实例

参数表装不下的几点

保存预览与重试

`tp submit`