LLM for finetuning to generate Urdu poema. Files are here: HERE Model is here: HERE Code is for Gemma and Llama. But Llama doesn't work with 8 vGPUs on ec2 so maybe we need much better GPUs