> For the complete documentation index, see [llms.txt](https://docs.zingg.ai/zingg/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.zingg.ai/zingg/running-zingg/create-training-data.md).

# Create Training Data

Run findTrainingData to generate candidate pairs from your data for labeling. The first active step in building your Zingg model.

The `findTrainingData` phase prompts Zingg to search for edge cases in your data-record pairs that are informative for learning. Zingg selects these judiciously so that your labeling effort is minimized and models can be built quickly.

Run `findTrainingData` first, then run the `label` phase. Repeat this cycle until you have enough labeled pairs.

{% hint style="success" icon="right-long" %}
Zingg selects the most informative pairs from your data, not random samples. Label until all field types and data variation patterns in your schema are represented. If accuracy needs improvement after the first match run, return to labeling and focus on patterns that are missing or underrepresented.
{% endhint %}

{% tabs %}
{% tab title="Community" %}

### Set label data sample size

```python
args.setLabelDataSampleSize(0.5)
```

### Run findTrainingData

```bash
./scripts/zingg.sh --phase findTrainingData --conf config.json
```

#### Using Python API

```python
options = ClientOptions([ClientOptions.PHASE, "findTrainingData"])
zingg = Zingg(args, options)
zingg.initAndExecute()
```

#### Using pre-existing training data

If you already have labeled pairs from a previous run or external source, you can supply them to Zingg using the `trainingSamples` attribute in your config.

Your training data must include:

* `z_cluster` - groups records together; uniquely identifies the group
* `z_isMatch` - `1` if the records match, `0` if they do not. Must be the same value for all records in the same `z_cluster` group.

#### Using JSON config

```json
{
  "trainingSamples": [
    {
      "name": "existingTraining",
      "format": "csv",
      "props": {
        "path": "/path/to/training.csv",
        "delimiter": ",",
        "header": "true"
      }
    }
  ]
}
```

{% hint style="success" icon="right-long" %}
Even when supplying pre-existing training data, it is advisable to run `findTrainingData` and `label` for a few rounds to tune Zingg with the patterns it needs to learn independently.
{% endhint %}
{% endtab %}

{% tab title="Enterprise" %}

### Set label data sample size

```python
args.setLabelDataSampleSize(0.5)
```

#### Python

```python
options = ClientOptions([ClientOptions.PHASE, "findTrainingData"])
zingg = EZingg(args, options)
zingg.initAndExecute()
```

#### CLI

```bash
./scripts/zingg.sh --phase findTrainingData --conf config.json
```

{% hint style="info" icon="right-long" %}
Candidate pairs are written to `zinggDir/modelId`. Run the label phase next to review and label these pairs.
{% endhint %}
{% endtab %}

{% tab title="Enterprise Snowflake" %}

### Run findTrainingData

Zingg on Snowflake can be run either from a local terminal via the CLI, or natively inside Snowflake as a job service.

#### CLI (local terminal)

```bash
./scripts/zingg.sh --phase findTrainingData --conf config.json \
--properties-file <location to snowflake.properties>
```

The `snowflake.properties` file supplies the connection details Zingg needs to reach your Snowflake account and run against your warehouse from outside Snowflake.

{% hint style="info" icon="right-long" %}
Candidate pairs are written to `zinggDir/modelId`.
{% endhint %}

#### Snowflake (job service)

Run the phase as an asynchronous job service inside your Snowflake compute pool.

```sql
EXECUTE JOB SERVICE
IN COMPUTE POOL CONTAINER_ZINGG_POOL
NAME = ZINGG_FTD_ASYNC_JOB_SERVICE
ASYNC = true
EXTERNAL_ACCESS_INTEGRATIONS = (ALLOW_ALL_EAI)
FROM @specs SPECIFICATION_TEMPLATE_FILE = '<zingg-job-spec-file>'
USING (PHASE => 'findTrainingData', CONFIG => '<config-name>');
```

The specification file passed via `SPECIFICATION_TEMPLATE_FILE` defines the container and job service configuration Zingg needs to run inside Snowpark Container Services. It is maintained by your administrator and referenced from your `@specs` stage.

`<config-name>` is the name of your Zingg configuration json file.

**Monitor the job**

Retrieve the service logs to monitor execution and debug any failures.

```sql
SELECT SYSTEM$GET_SERVICE_LOGS('ZINGG_FTD_ASYNC_JOB_SERVICE', 0, 'zingg-async-job-container');
```

`zingg-async-job-container` refers to the container name defined in the job specification.

**Stop the compute pool**

Once the job completes, stop the compute pool to release resources.

```sql
ALTER COMPUTE POOL CONTAINER_ZINGG_POOL STOP ALL;
```

{% endtab %}
{% endtabs %}


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.zingg.ai/zingg/running-zingg/create-training-data.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.