Update README.md

README.md

@@ -1,99 +1,7 @@
 ---
 license: apache-2.0
 base_model:
-- Qwen/Qwen3-0.6B
-library_name: transformers
-tags:
-- unsloth
-- reasoning
-- code
-- chain-of-thought
-- text-generation
-- shadow
-- conversational
-datasets:
-- unsloth/gsm8k
-- deepseek-ai/DeepSeek-R1
-pipeline_tag: text-generation
----
-
-# 🌑 Shadow 0.7B (Reasoning Edition)
-
-**Shadow 0.7B** is a specialized Small Language Model (SLM) optimized for **logical reasoning, competitive coding, and chain-of-thought processing**. 
-
-Built on the Qwen architecture and fine-tuned using **Unsloth**, Shadow punches far above its weight class, delivering "thinking" capabilities usually found in much larger models.
-
-## 🚀 Key Features
-* **🧠 Native Reasoning:** Trained to use `<think>` tags to plan and verify logic before answering.
-* **💻 Code Expert:** Optimized for Python and C++ algorithmic solutions (Chain of Draft).
-* **⚡ Lightweight:** Runs comfortably on free T4 GPUs, CPUs, and mobile devices (via Ollama).
-* **🆔 Custom Persona:** Maintains the identity of "Shadow", created by **Aman Kumar Pandey**.
-
-## 💻 Quick Start (Python)
-
-```python
-from transformers import AutoModelForCausalLM, AutoTokenizer
-
-model_name = "Redhanuman/Shadow-0.7B-Qwen3-Reasoning" # Replace with your actual username/repo
-
-model = AutoModelForCausalLM.from_pretrained(
-    model_name,
-    torch_dtype="auto",
-    device_map="auto"
-)
-tokenizer = AutoTokenizer.from_pretrained(model_name)
-
-# Shadow works best when you ask it to think
-prompt = "Write a Python script to check for palindromes. Explain your logic."
-messages = [
-    {"role": "user", "content": prompt}
-]
-
-text = tokenizer.apply_chat_template(
-    messages,
-    tokenize=False,
-    add_generation_prompt=True
-)
-
-model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
-
-generated_ids = model.generate(
-    **model_inputs,
-    max_new_tokens=1024
-)
-
-response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
-print(response)
-🦙 Run Locally (Ollama)
-If you have converted this model to GGUF, you can run it locally:
-
-Bash
-
-ollama run shadow
-🛠️ Training Details
-Creator: Aman Kumar Pandey (LPU)
-
-Framework: Unsloth (2x Faster Training)
-
-Base Model: Qwen 2.5 0.5B Instruct
-
-Method: QLoRA Fine-tuning with Chain of Draft (CoD) data.
-
-Created with ❤️ by Aman Kumar Pandey.
-
-
-### 📝 Instructions:
-1.  Go to your Model Page on Hugging Face.
-2.  Click **"Update model card"** (or create `README.md`).
-3.  **Delete everything** currently there.
-4.  **Paste** the code above.
-5.  **Important:** In the Python code section, make sure `Redhanuman/Shadow-0.7B-Qwen3-Reasoning` matches your *exact* repo name.
-6.  Click **Commit changes**.
-
-
----
-license: apache-2.0
-base_model: Qwen/Qwen2.5-0.5B-Instruct
+  - Qwen/Qwen3-0.6B
 library_name: transformers
 tags:
   - unsloth
@@ -111,22 +19,28 @@ pipeline_tag: text-generation
 
 # 🌑 Shadow 0.7B (Reasoning Edition)
 
-**Shadow 0.7B** is a specialized Small Language Model (SLM) optimized for **logical reasoning, competitive coding, and chain-of-thought processing**. 
+**Shadow 0.7B** is a specialized Small Language Model (SLM) optimized for **logical reasoning, competitive programming, and chain-of-thought processing**.
 
-Built on the Qwen architecture and fine-tuned using **Unsloth**, Shadow punches far above its weight class, delivering "thinking" capabilities usually found in much larger models.
+Built on the **Qwen3 0.6B** architecture and fine-tuned using **Unsloth**, Shadow delivers surprising reasoning depth and "thinking-first" responses uncommon for a model of this size.
+
+---
 
 ## 🚀 Key Features
-* **🧠 Native Reasoning:** Trained to use `<think>` tags to plan and verify logic before answering.
-* **💻 Code Expert:** Optimized for Python and C++ algorithmic solutions (Chain of Draft).
-* **⚡ Lightweight:** Runs comfortably on free T4 GPUs, CPUs, and mobile devices (via Ollama).
-* **🆔 Custom Persona:** Maintains the identity of "Shadow", created by **Aman Kumar Pandey**.
+
+* 🧠 **Structured Reasoning:** Uses `<think>` style internal reasoning patterns to improve answer quality.
+* 💻 **Coding Specialist:** Excels at Python, C++, and algorithmic problem-solving.
+* ⚡ **Ultra-Lightweight:** Runs on CPU, T4, mobile, or even low-VRAM consumer GPUs.
+* 🆔 **Custom Identity:** Retains the persona of **Shadow**, created by **Aman Kumar Pandey**.
+
+---
 
 ## 💻 Quick Start (Python)
 
 ```python
 from transformers import AutoModelForCausalLM, AutoTokenizer
+import torch
 
-model_name = "Redhanuman/Shadow-0.7B-Qwen3-Reasoning" # Replace with your actual username/repo
+model_name = "Redhanuman/Shadow-0.7B-Qwen3-Reasoning"  # Replace with your repo
 
 model = AutoModelForCausalLM.from_pretrained(
     model_name,
@@ -135,7 +49,6 @@ model = AutoModelForCausalLM.from_pretrained(
 )
 tokenizer = AutoTokenizer.from_pretrained(model_name)
 
-# Shadow works best when you ask it to think
 prompt = "Write a Python script to check for palindromes. Explain your logic."
 messages = [
     {"role": "user", "content": prompt}
@@ -147,37 +60,19 @@ text = tokenizer.apply_chat_template(
     add_generation_prompt=True
 )
 
-model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
+inputs = tokenizer([text], return_tensors="pt").to(model.device)
 
 generated_ids = model.generate(
-    **model_inputs,
+    **inputs,
     max_new_tokens=1024
 )
 
-response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
-print(response)
-🦙 Run Locally (Ollama)
-If you have converted this model to GGUF, you can run it locally:
-
-Bash
-
-ollama run shadow
-🛠️ Training Details
-Creator: Aman Kumar Pandey (LPU)
-
-Framework: Unsloth (2x Faster Training)
-
-Base Model: Qwen 2.5 0.5B Instruct
-
-Method: QLoRA Fine-tuning with Chain of Draft (CoD) data.
-
-Created with ❤️ by Aman Kumar Pandey.
-
+print(tokenizer.decode(generated_ids[0], skip_special_tokens=True))
+```
+## 🛠️ Training Details
 
-### 📝 Instructions:
-1.  Go to your Model Page on Hugging Face.
-2.  Click **"Update model card"** (or create `README.md`).
-3.  **Delete everything** currently there.
-4.  **Paste** the code above.
-5.  **Important:** In the Python code section, make sure `Redhanuman/Shadow-0.7B-Qwen3-Reasoning` matches your *exact* repo name.
-6.  Click **Commit changes**.
+- **Creator:** Aman Kumar Pandey (LPU)  
+- **Framework:** Unsloth (2× faster training)  
+- **Base Model:** Qwen3-0.6B  
+- **Method:** QLoRA fine-tuning with Chain-of-Draft (CoD) reasoning data  
+- **Datasets:** GSM8K, DeepSeek R1 distilled reasoning samples