> ## Documentation Index
> Fetch the complete documentation index at: https://docs.adaptive-ml.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Quickstart
> Fine-tune a model in 10 minutes
This guide walks you through training your first model on Adaptive Engine. You'll create a project, deploy a base model, upload training data, run fine-tuning, and test the results.
Train a customer service agent to respond like a poet.
Projects organize your models, data, and training runs.
Attach a model to deploy it for inference.
Datasets contain examples that teach your model how to respond.
Download: [fun-poet-completions.jsonl](https://huggingface.co/datasets/dylanebert/fun-poet-completions/resolve/main/fun-poet-completions.jsonl)
SFT (supervised fine-tuning) trains your model on labeled examples.
Deploy and chat with your fine-tuned model.
Set `ADAPTIVE_URL` and `ADAPTIVE_API_KEY` environment variables before starting. See [Authentication](/v0.14/advanced/authentication) for details.
Projects organize your models, data, and training runs.
```bash theme={null}
pip install adaptive_sdk
```
```python theme={null}
import os
from adaptive_sdk import Adaptive
adaptive = Adaptive(
base_url=os.environ["ADAPTIVE_URL"],
api_key=os.environ["ADAPTIVE_API_KEY"],
)
adaptive.projects.create(key="quickstart-fun-poet")
adaptive.set_default_project("quickstart-fun-poet")
```
Attach a model to deploy it for inference.
```python theme={null}
adaptive.models.attach(
model="qwen-3-4b-instruct-2507",
wait=True,
)
```
The model becomes available within a few minutes.
Datasets contain examples that teach your model how to respond.
Download: [fun-poet-completions.jsonl](https://huggingface.co/datasets/dylanebert/fun-poet-completions/resolve/main/fun-poet-completions.jsonl)
```python theme={null}
adaptive.datasets.upload(
file_path="fun-poet-completions.jsonl",
dataset_key="fun-poet-completions",
)
```
SFT (supervised fine-tuning) trains your model on labeled examples.
```python theme={null}
job = adaptive.jobs.run(
recipe_key="sft",
num_gpus=1,
args={
"model_to_train": "qwen-3-4b-instruct-2507",
"output_model_key": "fun-poet-v1",
"dataset": "fun-poet-completions",
"epochs": 3,
"samples_per_batch": 16,
"tp": 1,
},
)
print(f"Job started: {job.id}")
```
Training takes 15-30 minutes. Check progress in the UI under Runs.
Deploy and chat with your fine-tuned model.
```python theme={null}
adaptive.models.attach(
model="fun-poet-v1",
wait=True,
)
response = adaptive.chat.create(
model="fun-poet-v1",
messages=[
{"role": "user", "content": "I can't log into my account"}
],
)
print(response.choices[0].message.content)
```
The response should rhyme.
## Next steps
Rate responses to build training data
SFT, RLHF, DPO, and other methods