LFM2.5 Nova Function Calling

A compact 1.2B parameter liquid neural network model fine-tuned for function calling and tool use. This model excels at understanding tool descriptions, generating appropriate function calls, and handling multi-turn conversations with external API interactions.

Overview

LFM2.5-1.2B-Nova-Function-Calling is based on the Liquid Foundation Model architecture and has been specifically trained on the Nova-Synapse-Function-Calling dataset. Despite its small size, it demonstrates strong capabilities in structured output generation and tool orchestration, making it ideal for resource-constrained environments or edge deployments.

Key Features

**Function Calling**: Native support for generating structured function calls with proper parameter extraction

**Tool Use**: Can understand tool descriptions and select appropriate tools for given tasks

**Conversational**: Maintains context across multi-turn interactions while managing tool invocations

**Liquid Neural Network**: Uses dynamic, adaptive architecture for improved efficiency

**Lightweight**: At 1.2B parameters, runs efficiently on consumer hardware

**Multiple Quantizations**: Available in various GGUF quantization levels (Q2_K to f16) for different hardware constraints

Instructions

1. Choose and Download Quantization

Select a quantization level based on your hardware and quality requirements:

**Q4_K_M** (0.8 GB): Fast, recommended for most use cases

**Q5_K_M** (0.9 GB): Better quality with minimal speed impact

**Q6_K** (1.1 GB): Very good quality

**Q8_0** (1.3 GB): Best quality, still fast

Download from: https://huggingface.co/mradermacher/LFM2.5-1.2B-Nova-Function-Calling-GGUF

2. Load the Model

**Using llama.cpp:**

```bash

./main -m LFM2.5-1.2B-Nova-Function-Calling.Q4_K_M.gguf -n 512 -c 4096

```

**Using Ollama:**

```bash

Create a Modelfile

echo 'FROM ./LFM2.5-1.2B-Nova-Function-Calling.Q4_K_M.gguf' > Modelfile

ollama create nova-function-calling -f Modelfile

ollama run nova-function-calling

```

3. Define Your Tools

Provide clear tool descriptions in your prompt. Example format:

```

Available tools:

get_weather(location: string, units: string): Get current weather for a location

search_web(query: string, max_results: int): Search the web and return results

send_email(to: string, subject: string, body: string): Send an email

User: What's the weather in San Francisco?

```

4. Parse and Execute Function Calls

The model will generate structured function calls. Parse the output and execute the requested functions, then feed results back for continued conversation.

5. Best Practices

**Clear Tool Definitions**: Provide detailed parameter types and descriptions

**System Prompt**: Set context about available tools and expected output format

**Temperature**: Use 0.1-0.3 for more consistent function call generation

**Context Window**: Model supports up to 4096 tokens

**Error Handling**: Include error scenarios in your tool descriptions

Usage Example

```

System: You are a helpful assistant with access to tools. When you need to use a tool, output a function call in the format: FUNCTION_NAME(param1="value1", param2="value2")

User: Send an email to [email protected] about tomorrow's meeting

LFM2.5 Nova Function Calling

Safety Concern

LFM2.5 Nova Function Calling

Overview

Key Features

Instructions

1. Choose and Download Quantization

2. Load the Model

Create a Modelfile

3. Define Your Tools

4. Parse and Execute Function Calls

5. Best Practices

Usage Example

Reviews (0)