Update README.md

README.md

@@ -6,8 +6,6 @@ license: mit
 
 Making decoder-only transformers predict state consequences instead of tokens.
 
-🔗 **[View on Hugging Face](https://huggingface.co/wassemgtk/jepa_llm_prototypes)**
-
 ## What's This?
 
 Three approaches to convert a standard LLM into a world model that predicts "what happens next" given a state and action — like JEPA but for language models.
@@ -23,7 +21,7 @@ Three approaches to convert a standard LLM into a world model that predicts "wha
 ## Quick Start
 
 1. Open any notebook in [Google Colab](https://colab.research.google.com/)
-2. Set runtime to **GPU** (Runtime → Change runtime type → T4)
+2. Set runtime to **GPU** (Runtime → Change runtime type → H100)
 3. Run all cells
 4. Watch the model learn to predict state transitions
 
@@ -77,7 +75,7 @@ All dependencies install automatically in the notebooks.
 ## Next Steps
 
 - Swap synthetic data for real enterprise workflow logs
-- Scale up base model (Llama, Mistral)
+- Scale up base model (Llama, Qwen, Palmyra)
 - Add multi-step trajectory prediction
 - Integrate with planning/search algorithms