> ## Documentation Index
> Fetch the complete documentation index at: https://docs.platform.toloka.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# toloka training create

> Create a Training draft.

## Usage

```bash theme={null}
toloka training create [options] <name>
```

## Arguments

**name** `string` *required*

New Training name.

## Options

**--method \<method>** `string` (default: "sft")

Training method: sft or gisting.

**--base-model \<id>** `string`

Catalog Base Model source; pass exactly one of `--base-model` or `--model-version`.

**--model-version \<id>** `string`

Exact Model Version source; pass exactly one of `--base-model` or `--model-version`.

**--dataset \<name-or-id>** `string`

Dataset to train on: fills the empty cells of its split column (a new split column, 80/10/10, when it has none) and freezes a slice named after the Training; pass `--dataset` or `--slice`.

**--slice \<id>** `string`

Ready Dataset Slice that freezes a split column; pass `--dataset` or `--slice`.

**--split-column \<name-or-id>** `string`

Split column to train with, when the dataset has more than one.

**--parts \<names>** `string`

Train parts to learn on, comma-separated (default: every part with the train role).

**--lora-rank \<rank>** `integer`

SFT LoRA adapter rank: 8, 16, or 32 (default 16).

**--gist-tokens \<count>** `string`

Gisting soft tokens: 8, 32, or 64, or a fraction of the prompt "1/4", "1/8", "1/16", "1/32" (default "1/8").

**--epochs \<count>** `integer`

Training epochs, 1 through 5 (default 3).

**--learning-rate \<rate>** `number`

Optimizer learning rate (default 0.00015 for SFT, 0.03 for gisting).

**--batch-size \<count>** `integer`

Sequences per optimizer step: 8, 16, 32, or 64 (default 8).

**--max-sequence-length \<tokens>** `integer`

Token budget per example, 64 through 16384 (default 12288 for SFT, 3072 for gisting).

**--prompt-file \<path>** `string`

Required for gisting: non-empty UTF-8 prompt text file, at most 256 KiB.

**--json** `boolean` (default: false)

Print JSON.

## Inherited options

**-v, --version** `boolean`

Output the current version.

**--source-root \<directory>** `string`

Project root; otherwise search the working directory and parents; init and report use only the working directory.

**--api-key \<key>** `string`

API key override; login stores the key it verifies.

Environment fallback: [`TOLOKA_SCOPED_TOKEN`](https://docs.platform.toloka.ai/docs/cli/reference#environment-variables), [`TOLOKA_API_KEY`](https://docs.platform.toloka.ai/docs/cli/reference#environment-variables).

**--base-url \<url>** `string`

Platform base URL override; login saves the selected URL.

Environment fallback: [`TOLOKA_BASE_URL`](https://docs.platform.toloka.ai/docs/cli/reference#environment-variables).

**--project-id \<id>** `string`

Target project; overrides toloka.json.

Environment fallback: [`TOLOKA_PROJECT_ID`](https://docs.platform.toloka.ai/docs/cli/reference#environment-variables).

See the [CLI guide](https://docs.platform.toloka.ai/docs/cli.md#train) for the workflow, and [environment variables and exit codes](https://docs.platform.toloka.ai/docs/cli/reference.md#environment-variables) for shared behavior.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.