Update README.md

README.md

@@ -24,7 +24,7 @@ This model acts like a **startup co-pilot**, guiding users from rough ideas to c
 | Finetuned on     | ~1500 GPT-4-generated prompt-response pairs |
 | Format           | Instruction-tuned with `[INST] ... [/INST]` |
 | Finetuning method| QLoRA (4-bit), Flash Attention 2         |
-| Trainer          | Axolotl + RunPod (A6000 GPU)             |
+| Trainer          | Axolotl + RunPod (A40 GPU)             |
 
 ---
 
@@ -41,4 +41,27 @@ Training categories included:
 - Product planning
 - MVP prototyping
 - Marketing strategy
-- Startup operations
+- Startup operations
+
+---
+
+## Example Usage
+
+```python
+from transformers import AutoTokenizer, AutoModelForCausalLM
+
+model = AutoModelForCausalLM.from_pretrained("vitalune/ovarra-v1", device_map="auto")
+tokenizer = AutoTokenizer.from_pretrained("vitalune/ovarra-v1")
+
+prompt = "[INST] I want to launch an AI copywriting SaaS. Help me plan it. [/INST]"
+inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
+
+output = model.generate(**inputs, max_new_tokens=512)
+print(tokenizer.decode(output[0], skip_special_tokens=True))
+```
+
+## Credits
+
+- Fine-tuned and built by @vitalune
+- Based on prompt logic and data generation co-created using GPT-4 API
+- Hosted and accelerated via RunPod