Update README.md

README.md

@@ -1,31 +1,71 @@
 ---
+base_model: google/gemma-4-e4b-it
+library_name: transformers
 tags:
-- gguf
-- llama.cpp
 - unsloth
-- vision-language-model
+- instruction-tuning
+- aws
+- agentic-ai
+- lorA
+- conversational
 ---
 
-# gemma_4E4B-it_finetune : GGUF
+# Gemma-4-E4B-it Agentic AI on AWS (Instruct)
 
-This model was finetuned and converted to GGUF format using [Unsloth](https://github.com/unslothai/unsloth).
+This model is an instruction-tuned (Supervised Fine-Tuning) version of Google's [Gemma-4-E4B-it](https://huggingface.co/google/gemma-4-e4b-it). It has been specialized to act as a conversational assistant, answering complex architectural questions regarding Agentic AI systems, frameworks, and protocols on Amazon Web Services (AWS). 
 
-**Example usage**:
-- For text only LLMs:    `llama-cli -hf Capitaller/gemma_4E4B-it_finetune --jinja`
-- For multimodal models: `llama-mtmd-cli -hf Capitaller/gemma_4E4B-it_finetune --jinja`
+The model was fine-tuned using **[Unsloth](https://github.com/unslothai/unsloth)** to enhance its ability to reason about AWS architectures and provide actionable, structured guidance based on official AWS prescriptive documentation.
 
-## Available Model files:
-- `gemma-4-e4b-it.Q8_0.gguf`
-- `gemma-4-e4b-it.F16-mmproj.gguf`
+## Model Details
+* **Base Model:** `google/gemma-4-e4b-it` (via `unsloth/gemma-4-E4B-it`)
+* **Training Type:** Supervised Fine-Tuning (Instruction/Chat)
+* **Domain focus:** AWS Architecture, Agentic AI, Frameworks, and Protocols (MCP, etc.)
+* **Language:** English
+* **Library:** Unsloth / Hugging Face Transformers
 
-## ⚠️ Ollama Note for Vision Models
-**Important:** Ollama currently does not support separate mmproj files for vision models.
+## Dataset
+The model was trained on instruction-response pairs sourced from **AWS Prescriptive Guidance: Agentic AI frameworks, platforms, protocols, and tools on AWS**. It has been taught to answer queries concisely and provide highly technical, context-aware AWS architecture advice based on modern Agentic standards.
 
-To create an Ollama model from this vision model:
-1. Place the `Modelfile` in the same directory as the finetuned bf16 merged model
-3. Run: `ollama create model_name -f ./Modelfile`
-   (Replace `model_name` with your desired name)
+## Training Configuration
+Unlike a base model, this model already understands conversational flow. Fine-tuning was constrained to the attention and MLP layers to smoothly adapt its persona and specific technical knowledge without causing catastrophic forgetting.
 
-This will create a unified bf16 model that Ollama can use.
-This was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth)
-[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)
+* **Method:** PEFT / LoRA
+* **LoRA Rank (r):** 16 (Standard for Instruct tuning)
+* **LoRA Alpha:** 16 (or 32, scaled for optimal learning)
+* **Target Modules:** Attention and MLP modules (`q_proj`, `k_proj`, `v_proj`, `o_proj`, `gate_proj`, `up_proj`, `down_proj`)
+* **Precision:** 4-bit quantization (QLoRA) during training
+* **Optimizer:** Paged AdamW 8-bit
+
+## How to Use
+Because this is an Instruct model, you **must use the standard Gemma chat template** when querying it. You can interact with it exactly like a standard chatbot.
+
+You can load it using Transformers or Unsloth:
+
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+model_name = "Capitaller/gemma_4E4B-it_finetune"
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+model = AutoModelForCausalLM.from_pretrained(model_name)
+
+# Ensure you use the proper chat format
+messages = [
+    {"role": "user", "content": "How should I design an Agentic AI architecture on AWS that uses the Model Context Protocol (MCP)?"},
+]
+
+inputs = tokenizer.apply_chat_template(
+    messages,
+    add_generation_prompt=True,
+    tokenize=True,
+    return_tensors="pt"
+)
+
+outputs = model.generate(inputs, max_new_tokens=256, temperature=0.7)
+response = tokenizer.decode(outputs[0], skip_special_tokens=True)
+print(response)
+```
+
+### Prompting Tips
+This model is designed to be highly instructional and responsive to direct questions. 
+* Ask clear, technical questions: `"What are the recommended AWS compute platforms for hosting an MCP server?"`
+* Request specific structures: `"Write a brief step-by-step guide on securing an AI agent communication channel on AWS."`