Update README.md

README.md

@@ -1,67 +1,111 @@
+# 🧠 CBT-Copilot: LLaMA 3.2B Fine-Tuned for Cognitive Therapy
+
+Welcome to **CBT-Copilot**, an open-source LLM fine-tuned on therapy-aligned dialogues using the [Lumiiree/therapod-dpo](https://huggingface.co/datasets/Lumiiree/therapod-dpo) dataset. This model is designed to act as a **compassionate and supportive AI assistant**, trained in the tone of cognitive behavioral therapy (CBT), and suitable for mental health support applications.
+
 ---
-language: en
-license: apache-2.0
-library_name: transformers
-tags:
-- llama
-- causal-lm
-- merged
-- vllm
-inference:
-  parameters:
-    max_new_tokens: 256
-    temperature: 0.7
-    top_p: 0.9
-    repetition_penalty: 1.1
-datasets:
-- Lumiiree/therapod-dpo
-base_model:
-- meta-llama/Llama-3.2-3B-Instruct
+
+## 🔧 Model Details
+
+- **Base Model**: [`meta-llama/Llama-3.2-3B-Instruct`](https://huggingface.co/meta-llama/Llama-3.2-3B-Instruct)
+- **Fine-Tuning Method**: LoRA (Low-Rank Adaptation)
+- **Dataset**: [`Lumiiree/therapod-dpo`](https://huggingface.co/datasets/Lumiiree/therapod-dpo)
+- **Use Case**: Empathetic responses, journaling prompts, CBT-style thought reframing
+- **Trained by**: [Thillai Chithambaram](https://huggingface.co/thillaic)
+
 ---
 
-# CBT-Copilot 🧠
+## 🧠 Intended Use
+
+This model can be integrated into:
 
-CBT-Copilot is a fine-tuned version of `meta-llama/Llama-3.2-3B-Instruct`, designed to simulate conversations for cognitive behavioral therapy (CBT) support. It has been trained using LoRA and merged into a standalone model.
+- 💬 **Mental health chatbots**
+- 📔 **Journaling apps with AI reflections**
+- 🧘 **Self-help tools for cognitive restructuring**
+- 🧑‍⚕️ **Therapist assistants (non-clinical use)**
 
-The model is now compatible with `transformers`, `vLLM`, and other inference frameworks.
+> ⚠️ **Disclaimer**: This model is not a replacement for licensed mental health professionals. It should be used only as an assistant or for research.
 
-## 🚀 How to Use (vLLM)
+---
 
-You can serve it with [vLLM](https://github.com/vllm-project/vllm):
+## 🏗️ Training Configuration
 
-```bash
-python3 -m vllm.entrypoints.openai.api_server --model your-username/CBT-Copilot
+### ✅ LoRA Settings
+```python
+peft_config = LoraConfig(
+    r=8,
+    lora_alpha=16,
+    target_modules=["q_proj", "v_proj"],
+    lora_dropout=0.05,
+    bias="none",
+    task_type="CAUSAL_LM",
+)
 ```
 
-Then query it like this:
+### ✅ TrainingArguments
+```python
+args = TrainingArguments(
+    output_dir="llama-cbt-checkpoints",
+    per_device_train_batch_size=1,
+    gradient_accumulation_steps=4,
+    learning_rate=2e-5,
+    num_train_epochs=1,
+    logging_steps=100,
+    save_strategy="epoch",
+    bf16=True,
+    optim="paged_adamw_8bit",
+)
+```
+
+> Training was performed using Hugging Face's `transformers` + `peft` libraries with LoRA applied to key attention modules for lightweight adaptation.
+
+---
+
+## 🚀 How to Use
 
 ```python
-import openai
+from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
 
-openai.api_key = "EMPTY"
-openai.api_base = "http://localhost:8000/v1"
+model_id = "thillaic/CBT-Copilot"
+tokenizer = AutoTokenizer.from_pretrained(model_id)
+model = AutoModelForCausalLM.from_pretrained(model_id)
 
-response = openai.ChatCompletion.create(
-    model="CBT-Copilot",
-    messages=[
-        {"role": "user", "content": "I've been feeling really anxious lately. What can I do?"}
-    ]
-)
+pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)
+
+prompt = "I feel overwhelmed and stuck lately. What should I do?"
+response = pipe(prompt, max_new_tokens=200, do_sample=True, temperature=0.7)
 
-print(response["choices"][0]["message"]["content"])
+print(response[0]['generated_text'])
 ```
 
-## 🧠 Intended Use
+---
 
-This model is intended for educational and prototyping purposes in mental health-related chatbot systems. It is **not a substitute for professional therapy**.
+## 💡 Example Prompts
 
-## 📜 License
+- "I often feel like I’m not good enough. Help me reframe this thought."
+- "Give me a CBT-style journaling prompt for today."
+- "How can I deal with negative self-talk?"
+
+---
 
-This model is licensed under the Apache 2.0 license.
+## 🧾 License
 
+This project is open-sourced for educational and research purposes under the **MIT License**.
 
 ---
 
-*Model prepared and fine-tuned by **ThillaiC***
+## 🙏 Acknowledgements
+
+- Fine-tuned on the excellent [`therapod-dpo`](https://huggingface.co/datasets/Lumiiree/therapod-dpo) dataset
+- Built using Meta’s LLaMA 3.2B base model
+- LoRA integration powered by Hugging Face PEFT
+
+---
+
+## 🔗 Links
+
+- 🤗 Model: [huggingface.co/thillaic/CBT-Copilot](https://huggingface.co/thillaic/CBT-Copilot)
+- 📓 Dataset: [Lumiiree/therapod-dpo](https://huggingface.co/datasets/Lumiiree/therapod-dpo)
+
+---
 
----
+*Crafted with care by Thillai Chithambaram for the future of compassionate AI.*