Remove default system prompt, fix media_start token and update readme
#20
by bigmoyan - opened
- README.md +5 -0
- chat_template.jinja +1 -5
README.md
CHANGED
|
@@ -29,6 +29,11 @@ pipeline_tag: image-text-to-text
|
|
| 29 |
<b>📰 <a href="https://www.kimi.com/blog/kimi-k2-5.html">Tech Blog</a></b>
|
| 30 |
</p>
|
| 31 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 32 |
## 1. Model Introduction
|
| 33 |
|
| 34 |
Kimi K2.5 is an open-source, native multimodal agentic model built through continual pretraining on approximately 15 trillion mixed visual and text tokens atop Kimi-K2-Base. It seamlessly integrates vision and language understanding with advanced agentic capabilities, instant and thinking modes, as well as conversational and agentic paradigms.
|
|
|
|
| 29 |
<b>📰 <a href="https://www.kimi.com/blog/kimi-k2-5.html">Tech Blog</a></b>
|
| 30 |
</p>
|
| 31 |
|
| 32 |
+
## 0. Changelog
|
| 33 |
+
- 2026.1.29:
|
| 34 |
+
- The default system prompt might cause confusion to users and unexpected behaviours, so we remove it.
|
| 35 |
+
- The token `<|media_start|>` is incorrect; it has been replaced with `<|media_begin|>` in the chat template.
|
| 36 |
+
|
| 37 |
## 1. Model Introduction
|
| 38 |
|
| 39 |
Kimi K2.5 is an open-source, native multimodal agentic model built through continual pretraining on approximately 15 trillion mixed visual and text tokens atop Kimi-K2-Base. It seamlessly integrates vision and language understanding with advanced agentic capabilities, instant and thinking modes, as well as conversational and agentic paradigms.
|
chat_template.jinja
CHANGED
|
@@ -5,7 +5,7 @@
|
|
| 5 |
{%- elif c is not none -%}
|
| 6 |
{% for content in c -%}
|
| 7 |
{% if content['type'] == 'image' or content['type'] == 'image_url' -%}
|
| 8 |
-
<|
|
| 9 |
{% elif content['type'] == 'video' or content['type']== 'video_url'-%}
|
| 10 |
<|kimi_k25_video_placeholder|>
|
| 11 |
{% else -%}
|
|
@@ -57,10 +57,6 @@
|
|
| 57 |
<|im_system|>tool_declare<|im_middle|>{{ tools | tojson(separators=(',', ':')) }}<|im_end|>
|
| 58 |
{%- endif -%}
|
| 59 |
{%- endif -%}
|
| 60 |
-
|
| 61 |
-
{%- if messages|length == 0 or messages[0]['role'] != 'system' -%}
|
| 62 |
-
<|im_system|>system<|im_middle|>You are Kimi, an AI assistant created by Moonshot AI.<|im_end|>
|
| 63 |
-
{%- endif -%}
|
| 64 |
|
| 65 |
{%- for message in hist_msgs -%}
|
| 66 |
{{set_roles(message)}}
|
|
|
|
| 5 |
{%- elif c is not none -%}
|
| 6 |
{% for content in c -%}
|
| 7 |
{% if content['type'] == 'image' or content['type'] == 'image_url' -%}
|
| 8 |
+
<|media_begin|>image<|media_content|><|media_pad|><|media_end|>
|
| 9 |
{% elif content['type'] == 'video' or content['type']== 'video_url'-%}
|
| 10 |
<|kimi_k25_video_placeholder|>
|
| 11 |
{% else -%}
|
|
|
|
| 57 |
<|im_system|>tool_declare<|im_middle|>{{ tools | tojson(separators=(',', ':')) }}<|im_end|>
|
| 58 |
{%- endif -%}
|
| 59 |
{%- endif -%}
|
|
|
|
|
|
|
|
|
|
|
|
|
| 60 |
|
| 61 |
{%- for message in hist_msgs -%}
|
| 62 |
{{set_roles(message)}}
|