Update README.md

README.md

@@ -16,7 +16,31 @@ datasets:
 - nohurry/Opus-4.6-Reasoning-3000x-filtered
 - Jackrong/Qwen3.5-reasoning-700x
 ---
+# 🌟 Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled
+
+🔥 **Update (April 5): To help beginners and enthusiasts better understand and reproduce the fine-tuning process of this model, I have prepared the complete training notebook, codebase, and a comprehensive companion PDF guide! Please check the resource links below.**
+
+> ❤️ Special thanks to the Unsloth open-source library and @KyleHessling1 for their support.
+
+## 📚 Resources & Guides
+
+If you want to dive into how this model was trained, or wish to reproduce the results locally or on Colab, please visit my GitHub repository:
+👉 **[Jackrong-llm-finetuning-guide](https://github.com/R6410418/Jackrong-llm-finetuning-guide)**
+
+### 📥 Core Technical Document Direct Download
+You can click the link below to directly access the complete technical manual for the Qwopus3.5 training:
+
+* **[Qwopus3-5-27b-Colab_complete_guide_to_llm_finetuning.pdf](https://github.com/R6410418/Jackrong-llm-finetuning-guide/raw/main/Qwopus3-5-27b-Colab_complete_guide_to_llm_finetuning.pdf)**
+  * Covers the entire workflow, starting with an introduction to Google Colab and Unsloth.
+  * Details the complete pipeline with step-by-step explanations—from downloading the base model and normalizing heterogeneous data sources into a unified format, to configuring trainer hyperparameters and finally publishing to Hugging Face.
+  * Feedback is highly welcome! If you spot any shortcomings or areas for improvement, please let me know, and I will update it promptly.
 
+> **A Note:**
+> My goal in writing this guide goes beyond merely detailing a single training workflow. I want to convey a broader message: fine-tuning, post-training, and even medium-scale pre-training are not unattainable technical rituals, nor are they the exaggerated hype often packaged by social media. More often than not, all you need is a Google account, a standard laptop, and relentless curiosity. 
+> 
+> *No one starts as an expert. But every expert was once brave enough to begin.*
+
+---
 # 🌟 Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled
 
 > **Build Environment Upgrades:**
@@ -53,6 +77,8 @@ Let me analyze this request carefully:
             .
 ```
 
+---
+
 ## 🗺️ Training Pipeline Overview
 
 ```text
@@ -73,6 +99,7 @@ Final Model (Claude-4.6-Opus-Reasoning-Distilled,text-only)
 
 > **From the test results, it is clear that different Qwen3.5 quantized models show significant differences in tool-calling capability. Among them, only the 27B model distilled with Claude Opus reasoning demonstrates stable performance.**
 
+---
 
 🔥**Community-tested advantages** (benchmark tests by user @sudoing on a single RTX 3090):
 
@@ -91,6 +118,7 @@ Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled shows significant advantages in
 
  **Thanks to the community for the in-depth testing and feedback!**
 
+---
 
 ### 🔹 Supervised Fine-Tuning (SFT)
 - **Objective:** To inject high-density reasoning logic and establish a strict format for problem-solving involving an internal thinking state prior to outputting the final response.
@@ -103,7 +131,6 @@ The dataset consists of high-quality, filtered reasoning distillation data:
 | Dataset Name | Description / Purpose |
 |--------------|-----------------------|
 | [nohurry/Opus-4.6-Reasoning-3000x-filtered](https://huggingface.co/datasets/nohurry/Opus-4.6-Reasoning-3000x-filtered) | Provides comprehensive Claude 4.6 Opus reasoning trajectories. |
-| [TeichAI/claude-4.5-opus-high-reasoning-250x](https://huggingface.co/datasets/TeichAI/claude-4.5-opus-high-reasoning-250x) | Injecting high-intensity, structured reasoning instances. |
 | [Jackrong/Qwen3.5-reasoning-700x](https://huggingface.co/datasets/Jackrong/Qwen3.5-reasoning-700x) | Additional curated reasoning samples designed to strengthen structured step-by-step problem solving and improve reasoning diversity. |
 
 ## 🌟 Core Skills & Capabilities