Skip to main content

Fine-tuning API

The Fine-tuning API enables you to customize and train models for Named Entity Recognition (NER), text classification, and relation extraction tasks. Train models on your own datasets to achieve better performance for your specific use cases.

Available Endpoints​

Available Endpoints​

Open Models Catalog​

  • List Models - Retrieve a list of available open/base models and public adapters (GET /models/)
  • Get Model - Get details of a specific base model (GET /models/{model_id}/)

💡 Public Catalog: Endpoints under /models/ are publicly accessible and do not require authentication. They provide base models to choose from for fine-tuning.

Training & Fine-tuning Management​

  • Start Training - Create and start a new training job on a dataset (POST /trainings/)
  • List Trainings - Retrieve a paginated list of your custom fine-tuned models and training runs (GET /trainings/)
  • Get Training - Get details and status of a specific training job (GET /trainings/{training_id}/)
  • Delete Training - Delete a training job (DELETE /trainings/{training_id}/)
  • Stop Training - Stop a running training job (POST /trainings/{training_id}/stop/)

Training Monitoring & Evaluations​

  • List Events - List events for a training job (GET /trainings/{training_id}/events/)
  • Get Metrics - Get training metrics such as loss (GET /trainings/{training_id}/metrics/{metric}/)
  • Get Evaluations - Get evaluation metrics for a training job (GET /trainings/{training_id}/evaluations/)

Getting Started​

Training management endpoints (/trainings/...) require authentication with your API key:

HeaderDescription
KnowledgatorAPI-KeyYour API key
Content-Typeapplication/json

Example Usage​

SDK Usage​

💡 Dedicated SDK Guides: See Python SDK | TypeScript SDK

TypeScript/JavaScript​

import { createClient } from '@knowledgator/typescript-sdk';

const kg = createClient('YOUR_API_KEY');

// List available models
const models = await kg.finetuning.listModels();

// Start a training job
const training = await kg.finetuning.startTraining({
model_id: 'comprehend-it',
dataset_id: '6babd48a-2256-4466-b16b-9fd45ba59688',
hyperparameters: {
learning_rate: 0.001,
batch_size: 32
}
});

Python​

from knowledgator import Knowledgator

kg = Knowledgator(api_key="YOUR_API_KEY")

# List available models
models = kg.finetuning.list_models()

# Start a training job
training = kg.finetuning.start_training(
model_id="comprehend-it",
dataset_id="6babd48a-2256-4466-b16b-9fd45ba59688",
hyperparameters={
"learning_rate": 0.001,
"batch_size": 32
}
)

cURL​

# List available models (Public)
curl -X GET "https://api.platform.knowledgator.com/models/"

# Start a training job
curl -X POST "https://api.platform.knowledgator.com/trainings/" \
-H "KnowledgatorAPI-Key: YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model_id": "comprehend-it",
"dataset_id": "6babd48a-2256-4466-b16b-9fd45ba59688",
"hyperparameters": {
"learning_rate": 0.001,
"batch_size": 32
}
}'

Service Types​

The Fine-tuning API supports the following service types:

  • ner – Named Entity Recognition
  • text-classification – Text Classification
  • relation-extraction – Relation Extraction

Training Statuses​

Training jobs can have the following statuses:

  • pending – Training job created, waiting to start
  • running – Training is in progress
  • successful – Training completed successfully
  • failed – Training failed
  • stopped – Training was manually stopped

For detailed documentation on each endpoint, click on the links above.