v1.0.1

Author: thewebscraping

.github/workflows/ci.yml

@@ -33,21 +33,31 @@ jobs:
     needs: build
     runs-on: ubuntu-latest
     if: github.ref == 'refs/heads/main' && github.event_name == 'push'
+    strategy:
+      max-parallel: 1
+      matrix:
+        python-version: ['3.9']
     steps:
-      - uses: actions/checkout@v4
-      - name: Configure Git Credentials
-        run: |
-          git config user.name github-actions[bot]
-          git config user.email 41898282+github-actions[bot]@users.noreply.github.com
-      - uses: actions/setup-python@v5
-        with:
-          python-version: '3.10'
-      - run: echo "cache_id=$(date --utc '+%V')" >> $GITHUB_ENV
-      - uses: actions/cache@v4
-        with:
-          key: mkdocs-material-${{ env.cache_id }}
-          path: .cache
-          restore-keys: |
-            mkdocs-material-
-      - run: pip install -r requirements.txt
-      - run: mkdocs gh-deploy --force
+    - uses: actions/checkout@v4
+    - name: Set up Python ${{ matrix.python-version }}
+      uses: actions/setup-python@v5
+      with:
+        python-version: ${{ matrix.python-version }}
+    - name: Install Dependencies
+      run: |
+        python -m pip install --upgrade pip
+        pip install -r requirements-dev.txt
+    - name: Configure Git Credentials
+      run: |
+        git config user.name github-actions[bot]
+        git config user.email 41898282+github-actions[bot]@users.noreply.github.com
+    - uses: actions/cache@v4
+      with:
+        key: mkdocs-material-${{ env.cache_id }}
+        path: .cache
+        restore-keys: |
+          mkdocs-material-
+    - name: Publish Documentation
+      run: |
+        echo "cache_id=$(date --utc '+%V')" >> $GITHUB_ENV
+        mkdocs gh-deploy --force

.github/workflows/documentation.yml

@@ -0,0 +1,37 @@
+name: Documentation
+
+on:
+  release:
+    types: [created]
+
+jobs:
+  docs:
+    runs-on: ubuntu-latest
+    strategy:
+      max-parallel: 1
+      matrix:
+        python-version: ['3.9']
+    steps:
+    - uses: actions/checkout@v4
+    - name: Set up Python ${{ matrix.python-version }}
+      uses: actions/setup-python@v5
+      with:
+        python-version: ${{ matrix.python-version }}
+    - name: Install Dependencies
+      run: |
+        python -m pip install --upgrade pip
+        pip install -r requirements-dev.txt
+    - name: Configure Git Credentials
+      run: |
+        git config user.name github-actions[bot]
+        git config user.email 41898282+github-actions[bot]@users.noreply.github.com
+    - uses: actions/cache@v4
+      with:
+        key: mkdocs-material-${{ env.cache_id }}
+        path: .cache
+        restore-keys: |
+          mkdocs-material-
+    - name: Publish Documentation
+      run: |
+        echo "cache_id=$(date --utc '+%V')" >> $GITHUB_ENV
+        mkdocs gh-deploy --force

README.md

@@ -251,3 +251,5 @@ A new family of open language models demonstrating strong performance across aca
 * Google<end_of_turn>
 
 ```
+
+**Read the documentation:** [https://thewebscraping.github.io/gemma-template/](https://thewebscraping.github.io/gemma-template/)

data/test.json data/content-writer/test.jsondata/test.json renamed to data/content-writer/test.json

@@ -47,7 +47,7 @@ VMLU is a benchmark suite designed to evaluate foundational models with a focus
 |---------------------|:----------------:|:-----:|:--------------:|:----------:|:------:|:-----:|:----------:|
 | 1624257089558187281 | 05/01/2025 17:56 | 20.14 |     29.35      |   29.84    | 25.76  | 25.61 |    1497    |
 
-#### Results:
+#### Results
 * Out of 9,834 attempts, 1,497 responses were unanswered.
 * The dataset and evaluation results can be downloaded from the file: `gemma-benchmark/gemma_2b_vmlu_answers.csv`. Although it is not within the scope of this fine tuning.
 
@@ -57,20 +57,20 @@ VMLU is a benchmark suite designed to evaluate foundational models with a focus
 |---------------------|:----------------:|:-----:|:--------------:|:----------:|:------:|:-----:|:----------:|
 | 1840435368978448913 | 06/01/2025 19:04 | 36.11 |     43.45      |   41.92    | 39.06  | 39.64 |     82     |
 
-#### Results:
+#### Results
 * Out of 9,834 attempts, 82 responses were unanswered.
 * The dataset and evaluation results can be downloaded from the file: `gemma-benchmark/gemma_2b_it_vmlu_benchmark.csv`. Although it is not within the scope of this fine tuning.
 
-#### My Gemma Fine Tuning VMLU Score:
+#### My Gemma Fine Tuning VMLU Score
 
 ![Screenshot VMLU_Gemma_Fine_Tuning.png](images/Screenshot_VMLU_Gemma_Fine_Tuning.png)
 
-#### VMLU Leaderboard Score:
+#### VMLU Leaderboard
 There is a clear difference between the VMLU rankings in the Gemma 2B IT fine tuning, the score is close to the score of the **Gemma 7B IT** model. Here is a screenshot of the **VMLU Leaderboard** rankings:
 
 ![Screenshot VMLU_Gemma_Fine_Tuning.png](images/Screenshot_VMLU_Leaderboard.png)
 
-#### Additional Resources:
+#### Additional Resources
 * VMLU Website: [https://vmlu.ai/](https://vmlu.ai/)
 * VMLU Leaderboard: [https://vmlu.ai/leaderboard](https://vmlu.ai/leaderboard)
 * VMLU Github Repository: [https://github.com/ZaloAI-Jaist/VMLU/](https://github.com/ZaloAI-Jaist/VMLU/)

data/train.json data/content-writer/train.jsondata/train.json renamed to data/content-writer/train.json

@@ -1,7 +1,7 @@
 # Custom Templates to Vietnamese Language
 Gemma Template uses Jinja2 template.
 
-See also: [`models.Attr`](../../models/#attributes_5)
+See also: [`models.Attr`](../models.md#attr)
 
 * * *
 

docs/benchmark.md

@@ -1,7 +1,7 @@
 # Default Templates
 Gemma Template uses Jinja2 template.
 
-See also: [`models.Attr`](../../models/#attributes_5)
+See also: [`models.Attr`](../models.md#attributes_5)
 
 * * *
 

docs/custom_templates/custom_template.md

@@ -6,11 +6,13 @@
 True
 ```
 
-See also: [Method Arguments](../models/#method-arguments)
+See also: [Method Arguments](models.md#method-arguments)
 
 ## Generate User Prompt
 Create user prompt for Gemma Fine tuning.
 
+See also: [Method Arguments](models.md#method-arguments)
+
 !!! Parameters
     * **max_hidden_words (Union[int, float]):** default `0`.
         * Replace words in the document with '_____'.
@@ -105,6 +107,8 @@ Gemma open models are built from _____ same _____ and technology _____ Gemini mo
 ## Generate Model Prompt
 Create model prompt for Gemma Fine tuning.
 
+See also: [Method Arguments](models.md#method-arguments)
+
 ```pycon
 >>> prompt = gemma_template.generate_model_prompt(
 ...     document='This is a Test!',
@@ -134,6 +138,8 @@ Test
 ## Generate Prompt for Question
 Quickly create question prompts using the Gemma model.
 
+See also: [Method Arguments](models.md#method-arguments)
+
 ```pycon
 >>> prompt = gemma_template.generate_prompt(document='This is a Test!')
 '''

docs/custom_templates/default_template.md

@@ -3,7 +3,7 @@
 This library was developed for the Kaggle challenge:
 [**Google - Unlocking Global Communication with Gemma**](https://www.kaggle.com/competitions/gemma-language-tuning), sponsored by Google.
 
-## Credit Requirement
+### Credit Requirement
 
 **Important:** If you are a participant in the competition and wish to use this source code in your submission,
 you must clearly credit the original author before the competition's end date, **January 14, 2025**.
@@ -17,7 +17,7 @@ GitHub: [https://github.com/thewebscraping/gemma-template/](https://github.com/t
 LinkedIn: [https://www.linkedin.com/in/thetwofarm](https://www.linkedin.com/in/thetwofarm)
 ```
 
-# Overview
+## Overview
 
 Gemma Template is a lightweight and efficient Python library for generating templates to fine-tune models and craft prompts.
 Designed for flexibility, it seamlessly supports Gemma, LLaMA, and other language frameworks, offering fast, user-friendly customization.
@@ -35,64 +35,62 @@ As a newbie, I created Gemma Template based on what I read and learned from the
 
 Gemma Template supports exporting dataset files in three formats: `Text`, `Alpaca`, and `OpenAI`.
 
-# Multilingual Content Writing Assistant
+## Multilingual Content Writing Assistant
 
 This writing assistant is a multilingual professional writer specializing in crafting structured, engaging, and SEO-optimized content.
 It enhances text readability, aligns with linguistic nuances, and preserves original context across various languages.
 
 ---
 
-## Key Features:
-#### 1. **Creative and Engaging Rewrites**
+### Key Features:
+### 1. **Creative and Engaging Rewrites**
 - Transforms input text into captivating and reader-friendly content.
 - Utilizes vivid imagery and descriptive language to enhance engagement.
 
-#### 2. **Advanced Text Analysis**
+### 2. **Advanced Text Analysis**
 - Processes text with unigrams, bigrams, and trigrams to understand linguistic patterns.
 - Ensures language-specific nuances and cultural integrity are preserved.
 
-#### 3. **SEO-Optimized Responses**
+### 3. **SEO-Optimized Responses**
 - Incorporates keywords naturally to improve search engine visibility.
 - Aligns rewritten content with SEO best practices for discoverability.
 
-#### 4. **Professional and Multilingual Expertise**
+### 4. **Professional and Multilingual Expertise**
 - Full support for creating templates in local languages.
 - Supports multiple languages with advanced prompting techniques.
 - Vocabulary and grammar enhancement with unigrams, bigrams, and trigrams instruction template.
 - Supports hidden mask input text. Adapts tone and style to maintain professionalism and clarity.
 - Full documentation with easy configuration prompts and examples.
 
-#### 5. **Customize Advanced Response Structure and Dataset Format**
+### 5. **Customize Advanced Response Structure and Dataset Format**
 - Supports advanced response structure format customization.
 - Compatible with other models such as LLaMa.
 - Enhances dynamic prompts using Round-Robin loops.
 - Outputs multiple formats such as Text, Alpaca and OpenAI.
 
-**Installation**
-----------------
+## **Installation**
 
 To install the library, you can choose between two methods:
 
-#### **1\. Install via PyPI:**
+### **1\. Install via PyPI:**
 
 ```shell
 pip install gemma-template
 ```
 
-#### **2\. Install via GitHub Repository:**
+### **2\. Install via GitHub Repository:**
 
 ```shell
 pip install git+https://github.com/thewebscraping/gemma-template.git
 ```
 
-**Quick Start**
-----------------
+## **Quickstart**
 Start using Gemma Template with just a few lines of code:
 
-## Load Dataset
+### Load Dataset
 Returns: A Hugging Face Dataset or DatasetDict object containing the processed prompts.
 
-**Load Dataset from data dict**
+#### **Load Dataset from data dict**
 ```python
 from gemma_template import gemma_template
 
@@ -112,7 +110,8 @@ dataset = gemma_template.load_dataset(data_dict, output_format='text')   # enum:
 print(dataset['text'][0])
 ```
 
-**Load Dataset from local file path or HuggingFace dataset**
+#### **Load Dataset from local file path or HuggingFace dataset**
+
 ```python
 from gemma_template import gemma_template
 
@@ -133,7 +132,7 @@ dataset = gemma_template.load_dataset(
 )
 ```
 
-## Fully Customized Template
+### Fully Customized Template
 
 ```python
 from gemma_template import Template, FieldPosition, INPUT_TEMPLATE, OUTPUT_TEMPLATE, INSTRUCTION_TEMPLATE, PROMPT_TEMPLATE
@@ -169,7 +168,7 @@ response = template_instance.apply_template(
 print(response)
 ```
 
-### Output:
+#### Output
 
 ```text
 <start_of_turn>user