Chat Template

LFM2 uses a ChatML-like chat template to structure conversations. This page covers the chat template format for both text and vision models.

Overview

The chat template defines how conversations are structured using special tokens and roles. Here's an example conversation:

<|startoftext|><|im_start|>system
You are a helpful assistant trained by Liquid AI.<|im_end|>
<|im_start|>user
What is C. elegans?<|im_end|>
<|im_start|>assistant
It's a tiny nematode that lives in temperate soil environments.<|im_end|>

Format Structure

Special Tokens

Conversations are formatted using special tokens:

• <|im_start|> — Marks the beginning of a message. Always followed by the role name (system, user, assistant, or tool) and a line break.
• <|im_end|> — Marks the end of a message.
• <|startoftext|> — Optional token at the beginning of conversations (typically handled automatically).

Roles

LFM2 supports four conversation roles:

• system — (Optional) Sets the assistant's behavior and context. Defines who the assistant is and how it should respond.
• user — Messages from the user containing questions, instructions, or requests.
• assistant — Responses from the model.
• tool — Results from tool/function execution. Used for tool use workflows.

Text Models

For text-only LFM2 models, you can apply the chat template using the .apply_chat_template() method from Hugging Face Transformers:

from transformers import AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained("LiquidAI/LFM2.5-1.2B-Instruct")

messages = [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": "What is machine learning?"}
]

# Apply chat template
prompt = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)
print(prompt)

Vision Models

LFM2-VL vision-language models use the same ChatML-like format with additional support for images. When formatted, images are represented with a sentinel token (<image>), which is automatically replaced with image tokens by the processor.

Conversation Format

When creating conversations for vision models, use a structured format with content as a list containing image and text entries:

conversation = [
    {
        "role": "user",
        "content": [
            {"type": "image", "image": image},  # PIL Image or loaded image
            {"type": "text", "text": "What is in this image?"},
        ],
    },
]

Formatted Output

After applying the chat template, the conversation is formatted as:

<|startoftext|><|im_start|>system
You are a helpful multimodal assistant by Liquid AI.<|im_end|>
<|im_start|>user
<image>What is in this image?<|im_end|>
<|im_start|>assistant
This image shows a Caenorhabditis elegans (C. elegans) nematode.<|im_end|>

Usage

For vision models, you can apply the chat template using the processor's .apply_chat_template() method:

from transformers import AutoProcessor
from transformers.image_utils import load_image

processor = AutoProcessor.from_pretrained("LiquidAI/LFM2.5-VL-1.6B")

# Load image and create conversation
image = load_image("path/to/image.jpg")  # or use PIL Image
conversation = [
    {
        "role": "user",
        "content": [
            {"type": "image", "image": image},
            {"type": "text", "text": "What is in this image?"},
        ],
    },
]

# Apply chat template
prompt = processor.apply_chat_template(
    conversation,
    tokenize=False,
    add_generation_prompt=True
)
print(prompt)

Template Definition

The complete chat template definition can be found in the chat_template.jinja file in each model's Hugging Face repository.