Skip to content

Broken Formatting? [Gemma 4] #14

Description

@dakshcs
Image

using gemma4 e2b q4_k_m from unsloth, formatting works fine on other frontends.
for more info: using koboldcpp-x86_64-nocuda with custom mesa-git setup.

specific misrendered lines:

### Fork vs. Clone

It's important to distinguish between **forking** and **cloning**:

*   **Forking:** This is a **GitHub feature** that creates a *new repository* on GitHub under your account, copying the structure of the original.
*   **Cloning:** This is a **Git command** (`git clone <repository_url>`) that downloads a *copy* of a repository (either a public one or one you have access to) onto your local machine.
koboldcpp full logs

[SearchTool] supports_search_tools(gemma) = True
[SearchTool] supports_search_tools(gemma) = True
[ToolManager] get_enabled_tools_for_model(gemma) = {'text_get', 'search_memory', 'apply_text_edit'}
[CustomProvider] Using chat.completions API for model: gemma
[build_tools_for_provider] Built 1 tools for custom: ['search_memory']

Input: {"model": "gemma", "messages": [{"role": "system", "content": "You are a helpful assistant.\n\nWhen writing mathematical equations use LaTeX syntax with parentheses: \\( ... \\) for inline math and \\[ ... \\] for block math. You format your responses using markdown.\n\nYou can edit the currently selected target file using apply_text_edit. When a target file is selected, use target=\"file\". Use text_get with target=\"file\" to read the full text before editing.\n\nOperations (in order of preference):\n1. search_replace: Find exact text and replace it. Provide 'search' (text to find) and 'text' (replacement). The search must match EXACTLY including whitespace.\n2. replace: For small files, provide complete new content in 'text'.\n3. diff: Only for complex multi-location edits. Use valid unified diff format.\n\nExample search_replace:\n  operation: \"search_replace\"\n  search: \"def old_function():\n    pass\"\n  text: \"def new_function():\n    return True\"\n\nAlways provide a short summary of the change.\n\nYou have access to a search_memory tool that can search the user's past conversations and configured document directories for relevant context. Use this tool when:\n  - The user asks about something they mentioned before\n  - You need context from previous conversations\n  - The user references past discussions or decisions\n  - Finding relevant information from the user's documents would help\n\nThe search uses word-boundary matching, so searching for 'dog' will match 'dog', 'dog,', 'dog.' but not 'doggedly' or 'hotdog'. You can search 'history' (past conversations), 'documents' (configured directories), or 'all'.\n\nUse text_get to read the current text for a target before editing."}, {"role": "user", "content": "hi there! can you explain how github forks for me"}], "temperature": 1.0, "tools": [{"type": "function", "function": {"name": "search_memory", "description": "Search the user's past conversations and configured document directories for relevant context. Use this when the user asks about something they mentioned before, references past discussions, or when finding relevant information from their documents would help answer their question.", "parameters": {"type": "object", "properties": {"keyword": {"type": "string", "description": "The word or phrase to search for. Uses word-boundary matching, so 'dog' matches 'dog', 'dog,', 'dog.' but not 'doggedly'."}, "source": {"type": "string", "enum": ["history", "documents", "all"], "description": "Where to search: 'history' for past conversations, 'documents' for configured directories, or 'all' for both. Defaults to 'history'."}}}}}], "tool_choice": "auto"}

Processing Prompt (686 / 686 tokens)
Generating (54 / 300 tokens)
(EOS token triggered! ID:106)
[22:41:09] CtxLimit:740/8192, Init:0.02s, Processed:686 in 10.35s (66.31T/s), Generated:54/300 in 3.90s (13.86T/s), Total:14.26s
[TOOLCALL REASONING]: {
    "reasoning": "Explaining how GitHub forks works is general knowledge that I can provide directly without needing to search past conversations or documents.",
    "final_decision": "no",
    "tool_name": "null"
}

Processing Prompt (5 / 5 tokens)
Generating (605 / 1024 tokens)
(EOS token triggered! ID:106)
[22:41:52] CtxLimit:1099/8192, Init:0.00s, Processed:5 in 0.15s (32.47T/s), Generated:605/1024 in 42.51s (14.23T/s), Total:42.67s
Output: Here is an explanation of how GitHub forks work.

## What is a Fork in GitHub?

A **fork** on GitHub is essentially a personal copy of someone else's repository. When you fork a repository, you create a completely independent copy of the original repository under your own GitHub account.

This is a fundamental feature of Git and GitHub that allows developers to:

1.  **Experiment Safely:** You can take a project you like, modify it, or experiment with new features without affecting the original project's code or its maintainers.
2.  **Contribute Changes:** It provides a clean starting point for proposing changes back to the original project.
3.  **Create Independent Projects:** You can use an existing project as a foundation for a completely new project.

### How the Forking Process Works

The process is straightforward:

1.  **Find the Repository:** You navigate to the GitHub repository you want to work with.
2.  **Click "Fork":** On the main repository page, there is usually a "Fork" button (often in the upper right corner).
3.  **Create Copy:** When you click "Fork," GitHub creates a new repository under your GitHub account, containing all the files, branches, and history of the original repository at that moment.

### Fork vs. Clone

It's important to distinguish between **forking** and **cloning**:

*   **Forking:** This is a **GitHub feature** that creates a *new repository* on GitHub under your account, copying the structure of the original.
*   **Cloning:** This is a **Git command** (`git clone <repository_url>`) that downloads a *copy* of a repository (either a public one or one you have access to) onto your local machine.

**In summary:** You fork on GitHub to create a *remote copy* on the platform. Then, you typically **clone** your fork to your local computer so you can start making changes.

### Workflow After Forking

Once you have forked a repository, the typical workflow for contributing back to the original project is:

1.  **Clone Your Fork Locally:** You clone your new fork to your local machine.
2.  **Make Changes:** You make the desired edits, add new features, or fix bugs in your local copy.
3.  **Commit Changes:** You commit your changes to a new branch in your local repository.
4.  **Push to Your Fork:** You push those changes to your repository on GitHub (your fork).
5.  **Create a Pull Request (PR):** You then go to the *original* repository on GitHub and open a "Pull Request," asking the original maintainers to review and potentially merge your changes into their project.

Let me know if you'd like a deeper dive into any specific step, like how to create a Pull Request!

(ChatGTK.py:58150): Gtk-WARNING **: 22:41:52.594: Invalid markup string: Value of 'background' attribute on <span> tag on line 2 could not be parsed; should be a color specification, not 'rgba(89,174,249,0.25)'

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions