finetune: update the params to run experiment on H200

fix: reduce templates to country, region & county for mvp

enh: Run gazet on modal - llama server on GPU, API & UI on cpu

Lowercase subtype names, fix question type hints

Fix: No pairs are created for mixed queries

prompt: Add notes on how to pick places from query better

Fix natural earth geometries by converting wkb geometries to geojson

chore: clean sql generation, use conversation format, move prompts from user to system

Add model training script for qwen, update eval & gazet lm to work with qwen template

Add finetune instructions for gemma

Add options to run gguf models using llama cpp

Revert ollama port to 11434

ENH: Add dataset creation on scale using modal clusters

ENH: use separate models for places & sql generation

Disable dspy cache

Intitial commit