Change code tags to xml #1442
Change code tags to xml #1442
Conversation
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
| @@ -205,7 +198,7 @@ planning: | |||
| Then for the given task, develop a step-by-step high-level plan taking into account the above inputs and list of facts. | |||
| This plan should involve individual tasks based on the available tools, that if executed correctly will yield the correct answer. | |||
| Do not skip steps, do not add any superfluous steps. Only write the high-level plan, DO NOT DETAIL INDIVIDUAL TOOL CALLS. | |||
| After writing the final step of the plan, write the '\n<end_plan>' tag and stop there. | |||
| After writing the final step of the plan, write the '<end_plan>' tag and stop there. | |||
There was a problem hiding this comment.
Why not aligning this tag as well, and use </plan> instead of <end_plan>?
There was a problem hiding this comment.
Here since there's no opening tag , maybe alone would be strange?
The idea of <end_plan> is to cut generation and prevent the model from doing anything else after generating its plan.
| @@ -180,6 +168,11 @@ system_prompt: |- | |||
| 9. The state persists between code executions: so if in one step you've created variables or imported modules, these will all persist. | |||
| 10. Don't give up! You're in charge of solving the task, not providing directions to solve it. | |||
|
|
|||
| {%- if custom_instructions %} | |||
| Here are custom instructions: | |||
There was a problem hiding this comment.
Here are custom instructions:
Does this sentence add value to the prompt? Does the LLM treat differently "custom instructions" versus "regular"(?) instructions?
There was a problem hiding this comment.
Good point, I've removed this sentence since it adds little value.
| assert "```python" in tool_message.content | ||
| else: | ||
| assert expected in tool_message.content | ||
| assert expected in tool_message.content |
There was a problem hiding this comment.
Removing the if/else here since I didn't see why it was needed!
Co-authored-by: Albert Villanova del Moral <8515462+albertvillanova@users.noreply.github.com>
Change code tags to XML to align with the widespread usage of XML tags within chat template for most LLMs.
For instance gemma-3 chat template uses XML formatting: check it out here. Many other models also use XML: Qwen for instance.
Testing XML tags on models like GPT-4o, Claude-4-Sonnet, Llama-3.1 8B & 70B, or Mistral 7B, didn't show any performance degradation.