> ## Documentation Index > Fetch the complete documentation index at: https://docs.valkyrie.vals.ai/llms.txt > Use this file to discover all available pages before exploring further. # valkyrie dataset > Complete generated reference for every `valkyrie dataset` command. This page documents all 13 commands in the `dataset` group in registration order. Use the table of contents to jump to a command. ## `valkyrie dataset create` Create a dataset in a benchmark. BENCHMARK must already exist. `valkyrie dataset push` creates a missing one. NAME is lowercase letters and digits with single '.', '\_' or '-' separators, such as verified. ```bash Syntax theme={null} valkyrie dataset create \ \ \ [--description ] ``` ```bash Example theme={null} valkyrie dataset create \ swe-bench \ verified \ --description "Human-verified issues" ``` **Arguments** BENCHMARKrequired · `text` NAMErequired · `text` **Options** \--descriptionoptional · `text` · default: `''` What the dataset holds. Task guidance: [Dataset guides](https://docs.valkyrie.vals.ai/benchmarks/datasets.md). *** ## `valkyrie dataset list` List the datasets in a benchmark. ```bash Syntax theme={null} valkyrie dataset list \ \ [--include-archived] \ [--format ] ``` **Arguments** BENCHMARKrequired · `text` **Options** \--include-archivedoptional · `boolean` · default: `False` Also list archived ones, with when each was archived. \--formatoptional · `choice` · default: `'text'` · constraints: one of text, json; case-insensitive Output format. Task guidance: [Dataset guides](https://docs.valkyrie.vals.ai/benchmarks/datasets.md). *** ## `valkyrie dataset show` Show one dataset, archived or not, with its latest version and the dataset.toml it was published with. ```bash Syntax theme={null} valkyrie dataset show \ \ \ [--format ] ``` **Arguments** BENCHMARKrequired · `text` NAMErequired · `text` **Options** \--formatoptional · `choice` · default: `'text'` · constraints: one of text, json; case-insensitive Output format. Task guidance: [Dataset guides](https://docs.valkyrie.vals.ai/benchmarks/datasets.md). *** ## `valkyrie dataset update` Change a dataset's description. Its name never changes. ```bash Syntax theme={null} valkyrie dataset update \ \ \ --description ``` ```bash Example theme={null} valkyrie dataset update \ swe-bench \ verified \ --description "SWE-bench Verified" ``` **Arguments** BENCHMARKrequired · `text` NAMErequired · `text` **Options** \--descriptionrequired · `text` The description the dataset now has. Task guidance: [Dataset guides](https://docs.valkyrie.vals.ai/benchmarks/datasets.md). *** ## `valkyrie dataset remove` Archive a dataset. It stays readable, keeps its name, and can be restored. ```bash Syntax theme={null} valkyrie dataset remove \ \ \ [--yes|-y] ``` **Arguments** BENCHMARKrequired · `text` NAMErequired · `text` **Options** \--yes, -yoptional · `boolean` · default: `False` Skip the confirmation prompt. Task guidance: [Dataset guides](https://docs.valkyrie.vals.ai/benchmarks/datasets.md). *** ## `valkyrie dataset restore` Restore an archived dataset so its tasks can change again. ```bash Syntax theme={null} valkyrie dataset restore \ \ ``` **Arguments** BENCHMARKrequired · `text` NAMErequired · `text` Task guidance: [Dataset guides](https://docs.valkyrie.vals.ai/benchmarks/datasets.md). *** ## `valkyrie dataset push` Publish the task folders under FOLDER as a new version of a dataset. Only new tasks are uploaded. ```bash Syntax theme={null} valkyrie dataset push \ \ \ [--tag ...] \ [--concurrency ] ``` **Arguments** BENCHMARK/DATASETrequired · `text` FOLDERrequired · `path` · constraints: must exist; directory; readable **Options** \--tagoptional · `text` · constraints: repeatable Also tag the version. Repeatable. \--concurrencyoptional · `integer` · default: `4` · constraints: >= 1 Tasks to upload at once Task guidance: [Dataset guides](https://docs.valkyrie.vals.ai/benchmarks/datasets.md). *** ## `valkyrie dataset install` Download a dataset from Harbor, GitHub or Hugging Face and publish it as a new version. ```bash Syntax theme={null} valkyrie dataset install \ \ \ --location \ [--tag ...] \ [--concurrency ] ``` ```bash Example theme={null} valkyrie dataset install \ tb/tb2 \ terminal-bench/terminal-bench-2 \ --location harbor ``` **Arguments** BENCHMARK/DATASETrequired · `text` SOURCErequired · `text` **Options** \--locationrequired · `choice` · constraints: one of harbor, github, hugging\_face Where SOURCE is. \--tagoptional · `text` · constraints: repeatable Also tag the version. Repeatable. \--concurrencyoptional · `integer` · default: `4` · constraints: >= 1 Tasks to upload at once Task guidance: [Dataset guides](https://docs.valkyrie.vals.ai/benchmarks/datasets.md). *** ## `valkyrie dataset versions` List the versions of a dataset, newest first, with the tags on each. ```bash Syntax theme={null} valkyrie dataset versions \ \ [--format ] ``` **Arguments** BENCHMARK/DATASETrequired · `text` **Options** \--formatoptional · `choice` · default: `'text'` · constraints: one of text, json; case-insensitive Output format. Task guidance: [Dataset guides](https://docs.valkyrie.vals.ai/benchmarks/datasets.md). *** ## `valkyrie dataset tag` Point a tag at a version of a dataset. The version is the latest one unless the reference ends in @ and a revision number, a tag or a digest. A tag that points at another version moves only with --force. The latest tag always moves. A tag is lowercase letters, digits, hyphens and dots, starts with a letter or digit, and can't be a whole number or start with sha256:. ```bash Syntax theme={null} valkyrie dataset tag \ \ \ [--force] ``` ```bash Example theme={null} valkyrie dataset tag \ swe-bench/verified@2 \ stable ``` **Arguments** BENCHMARK/DATASET\[@VERSION]required · `text` TAGrequired · `text` **Options** \--forceoptional · `boolean` · default: `False` Move the tag when it already points at another version. Task guidance: [Dataset guides](https://docs.valkyrie.vals.ai/benchmarks/datasets.md). *** ## `valkyrie dataset untag` Remove a tag from a dataset. The latest tag can't be removed. ```bash Syntax theme={null} valkyrie dataset untag \ \ ``` **Arguments** BENCHMARK/DATASETrequired · `text` TAGrequired · `text` Task guidance: [Dataset guides](https://docs.valkyrie.vals.ai/benchmarks/datasets.md). *** ## `valkyrie dataset task list` List the tasks of a dataset version: the latest one, or the version after @, which is a revision number, a tag or a digest. ```bash Syntax theme={null} valkyrie dataset task list \ \ [--format ] ``` ```bash Example theme={null} valkyrie dataset task list swe-bench/verified@2 ``` **Arguments** BENCHMARK/DATASET\[@VERSION]required · `text` **Options** \--formatoptional · `choice` · default: `'text'` · constraints: one of text, json; case-insensitive Output format. Task guidance: [Dataset guides](https://docs.valkyrie.vals.ai/benchmarks/datasets.md). *** ## `valkyrie dataset task download` Download a task's archive from the latest version of a dataset, or from the version after @. It is saved as TASK\_ID.zip unless you pass --output. ```bash Syntax theme={null} valkyrie dataset task download \ \ \ [--output|-o ] \ [--overwrite] ``` **Arguments** BENCHMARK/DATASET\[@VERSION]required · `text` TASK\_IDrequired · `text` **Options** \--output, -ooptional · `path` · constraints: file; readable Where to save the archive. \--overwriteoptional · `boolean` · default: `False` Replace the file if it exists. Task guidance: [Dataset guides](https://docs.valkyrie.vals.ai/benchmarks/datasets.md). This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.