GPT4All V3.6.1->Jinia error

[1015g]

jq -r ".chat_template" tokenizer_config.json

Jinja out of the box in GPT4All:

{% for message in messages %}{{'<|im_start|>' + message['role'] + '
' + message['content']}}{% if not loop.last or (loop.last and message['role'] != 'assistant') %}{{'<|im_end|>
'}}{% endif %}{% endfor %}{% if add_generation_prompt and messages[-1]['role'] != 'assistant' %}{{ '<|im_start|>assistant
' }}{% endif %}

(#3340)

[1015g]

Sorry, I don't know how to modify it according to the prompts. Because I don't understand anything.And that's just one of my many local models that have problems., which is terrible! Can't believe this could be an updated change!

[ThiloteE]

As long as you do not provide a link to the model you are using, nobody can reproduce and test an improved chat template, but simply by looking at the jinja, the following could work. Please try:

{%- for message in messages %}
    {{- '<|im_start|>' + message['role'] + '\n' + message['content'] }}
    {%- if not loop.last or (loop.last and message['role'] != 'assistant') %}
        {{- '<|im_end|>\n' }}
    {%- endif %}
{%- endfor %}
{%- if add_generation_prompt and messages[-1]['role'] != 'assistant' %}
    {{- '<|im_start|>assistant\n' }}
{%- endif %}

[1015g]

As long as you do not provide a link to the model you are using, nobody can reproduce and test an improved chat template, but simply by looking at the jinja, the following could work. Please try:

{%- for message in messages %}
{{- '<|im_start|>' + message['role'] + '\n' + message['content'] }}
{%- if not loop.last or (loop.last and message['role'] != 'assistant') %}
{{- '<|im_end|>\n' }}
{%- endif %}
{%- endfor %}
{%- if add_generation_prompt and messages[-1]['role'] != 'assistant' %}
{{- '<|im_start|>assistant\n' }}
{%- endif %}

Thank you very much for your help! Here is the model link:https://huggingface.co/tastypear/CausalLM-7B-DPO-alpha-GGUF