Gemma inference with llama.cpp and logits processor

scripts/local.sh

@@ -27,6 +27,7 @@ docker run \
     -ti \
     --gpus=all \
     -p 8080:8080 \
+    -p 8000:8000 \
     --env-file .env \
     -v /$(pwd)/models/:/root/models/ \
     -v /$(pwd)/iterations/:/root/data/iterations/ \