LFM2.5 Nova Function Calling

A lightweight 1.2B parameter liquid neural network model fine-tuned for function calling and tool use in conversational contexts. This GGUF quantized model provides efficient function execution capabilities while maintaining a small memory footprint.

Description

LFM2.5-1.2B-Nova-Function-Calling is based on NovachronoAI's liquid neural network architecture and trained on the Nova-Synapse-Function-Calling dataset. The model excels at:

This skill provides weighted/imatrix GGUF quantizations optimized for various performance/quality tradeoffs, from ultra-compressed (IQ1_S ~0.4GB) to high-quality (Q6_K ~1.1GB) variants.

Instructions

1. Model Selection

Choose a quantization based on your requirements:

**Recommended for most use cases:**

**For limited resources:**

**For maximum quality:**

**For extreme compression (not recommended):**

2. Download and Load

Download your chosen quantization from the model repository:

```bash

Example: Download Q4_K_M variant

wget https://huggingface.co/mradermacher/LFM2.5-1.2B-Nova-Function-Calling-i1-GGUF/resolve/main/LFM2.5-1.2B-Nova-Function-Calling.i1-Q4_K_M.gguf

```

Load with llama.cpp:

```bash

./main -m LFM2.5-1.2B-Nova-Function-Calling.i1-Q4_K_M.gguf \

--ctx-size 4096 \

--temp 0.7 \

--repeat-penalty 1.1

```

Or with Ollama:

```bash

Create Modelfile

FROM ./LFM2.5-1.2B-Nova-Function-Calling.i1-Q4_K_M.gguf

PARAMETER temperature 0.7

PARAMETER stop "<|im_end|>"

Import and run

ollama create lfm-nova -f Modelfile

ollama run lfm-nova

```

3. Function Calling Format

Structure your prompts to include function definitions and user requests:

```

Available functions:

User: What's the weather like in Paris?