Local LLMs

**Running DeepSeek R1 in LLM Studio**

1. **Install LLM Studio (v0.12+)**
“`bash
pip install llm-studio
“`

2. **Add DeepSeek R1 model**
“`bash
llm studio add-model deepseek-ai/DeepSeek-Coder-1.3B-Instruct
“`

3. **Launch the UI**
“`bash
llm studio start
“`
Open `http://localhost:8080` in your browser.

4. **Configure the model**
– In the sidebar, select *DeepSeek R1*.
– Set `max_tokens`, `temperature`, and other inference params as desired.

5. **Run a prompt**
“`text
# Prompt
def hello_world():
print(“Hello, world!”)
“`
Click *Run* and watch the instant response.

6. **Fine‑tune (optional)**
“`bash
llm studio finetune deepseek-ai/DeepSeek-Coder-1.3B-Instruct –dataset my_dataset.jsonl
“`

**Tips**

| Tip | Why |
|—–|—–|
| Use `–max_new_tokens 512` | Avoids excessive generation time |
| Enable *GPU Acceleration* in settings | Drastically speeds up inference |
| Save prompts as templates | Reuse common coding patterns |

That’s it—DeepSeek R1 is now running locally in LLM Studio, ready for rapid prototyping and experimentation. Happy coding!