chat_template.jinja

{{ bos_token }}
{%- set prefix = """You are a safety classifier for analyzing interactions between humans and LLMs. Given a request from a human user and a response from an LLM assistant, provide a safe or unsafe 'rating' for both the user input and the assistant response. 
Taxonomy: The following are the list of harmful categories that the input and response may fall into.
<BEGIN UNSAFE CONTENT CATEGORIES>
S1: Violence.
S2: Sexual.
S3: Criminal Planning/Confessions.
S4: Guns and Illegal Weapons.
S5: Controlled/Regulated Substances.
S6: Suicide and Self Harm.
S7: Sexual (minor).
S8: Hate/Identity Hate.
S9: PII/Privacy.
S10: Harassment.
S11: Threat.
S12: Profanity.
S13: Needs Caution.
S14: Other.
S15: Manipulation.
S16: Fraud/Deception.
S17: Malware.
S18: High Risk Gov Decision Making.
S19: Political/Misinformation/Conspiracy.
S20: Copyright/Trademark/Plagiarism.
S21: Unauthorized Advice.
S22: Illegal Activity.
S23: Immoral/Unethical.
<END UNSAFE CONTENT CATEGORIES>
<BEGIN CONVERSATION>
user: """
-%}

{%- set suffix = """<END CONVERSATION>
Use the taxonomy to determine the safety of the user input and assistant response. Output your answer in the following format:
User Safety: safe or unsafe rating of the user input
Response Safety: safe or unsafe rating of the assistant response. Omit if no assistant response present.
Safety Categories: a comma-separated list of applicable safety categories from the provided taxonomy. Omit if all safe or not needed.

Do not include anything other than the output in the above format in your response. """ ~ (request_categories or "/no_categories") ~ "\nOutput:" 
-%}

{%- set ns = namespace(prompt='', image_token='') -%}

{%- if messages[0]['role'] == 'system' -%}
    {%- if messages[0]['content'] is string -%}
        {%- set first_user_prefix = messages[0]['content'] + '\n\n' -%}
    {%- else -%}
        {%- set first_user_prefix = messages[0]['content'][0]['text'] + '\n\n' -%}
    {%- endif -%}
    {%- set loop_messages = messages[1:] -%}
{%- else -%}
    {%- set first_user_prefix = "" -%}
    {%- set loop_messages = messages -%}

{%- endif -%}

{{ "<start_of_turn>user\n" }}
{%- for message in loop_messages -%}
    {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
        {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
    {%- endif -%}
    {%- if (message['role'] == 'assistant') -%}
        {%- set role = "model" -%}
    {%- else -%}
        {%- set role = message['role'] -%}
    {%- endif -%}
    
    {%- if message['content'] is string -%}
        {{ message['content'] | trim }}
    {%- elif message['content'] is iterable -%}
        {%- for item in message['content'] -%}
            {%- if item['type'] == 'image' -%}
                {%- if loop.last -%}
                    {{ '<start_of_image>' + ns.prompt | trim }}
                {%- else -%}
                    {%- set ns.image_token="<start_of_image>" -%}
                {%- endif -%}
            {%- elif item['type'] == 'text' -%}
                {%- if (message['role'] == 'user') -%}
                    {%- if loop.last -%}
                        {{ ns.image_token + prefix + item['text'] | trim }}
                    {%- else -%}
                        {%- set ns.prompt = prefix + item['text'] | trim -%}
                    {%- endif -%}

                {%- else -%}
                    {{ 'response: agent:  ### Answer: ' + item['text'] | trim }}
                {%- endif -%}
                
            {%- endif -%}
        {%- endfor -%}
    {%- else -%}
        {{ raise_exception("Invalid content type") }}
    {%- endif -%}
    {{ '\n' }}

{%- endfor -%}

{{ suffix + "<end_of_turn>\n<start_of_turn>model\n" }}