update readme for llm-transformers example

examples/llm-transformers/README.md

@@ -61,3 +61,58 @@ To get the path, activate the environment and execute `which python` (linux) or
 The path to the transformers cache (where all the models are stored) can be changed with `transformerscache` option.
 Leave it out completely to used the default (usually in home folder).
 The sized for `70B` variants (one models) are usually around 130 GB.
+
+
+## OAEI
+
+### Build the docker image
+To create the OAEI matcher, uncomment the corresponding build module in the `pom.xml` file.
+Then run `mvn install`.
+
+If you want to use `podman` instead of `docker` (only during build!) you need to run 
+`podman system service {dashdash}time=0 unix:/run/user/$(id -u)/podman/podman.sock`
+in a seperate shell and then before executing maven run `export DOCKER_HOST="unix:/run/user/$(id -u)/podman/podman.sock" `.
+
+### Execute the docker image directly (without MELT)
+Running this matcher in a reasonable time requires GPUs (in experiments we used 2 x A100 with 40 GB RAM).
+
+To execute the image run, 
+
+```
+docker run --rm --publish 8080:8080 llm-transformers-1.0-web
+```
+(or replace `docker` by `podman`).
+
+This will download the large language model (~250GB) into the container which means that if you run the container again, it will be downloaded again.
+To map the cache folder to a folder on your hard drive (replace `{localPath}` with it) you need to execute:
+
+```
+docker run --rm --publish 8080:8080 -v {localPath}:/root/.cache/ llm-transformers-1.0-web
+```
+
+With this setup it only downloads the large language model only once even if the container is started multiple times.
+
+If a restriction on specific GPUs is required, set the environemnt variable `CUDA_VISIBLE_DEVICES` like
+
+```
+docker run --rm --publish 8080:8080 -e CUDA_VISIBLE_DEVICES=0 llm-transformers-1.0-web
+```
+
+Both options can also be combined.
+
+Once the container is started, it will listen on the 8080 port and provides a 
+REST API defined by the [MELT Web format](https://dwslab.github.io/melt/matcher-packaging/web).
+
+### Execute the docker image with MELT
+
+[Download the MELT evaluation client](https://dwslab.github.io/melt/matcher-evaluation/client) and run:
+
+```
+java -jar matching-eval-client-latest.jar --systems <path to the tar.gz file> --track <location-URI> <collection-name> <version>
+```
+
+if the docker container is already running you need to provide the locahost URL:
+
+```
+java -jar matching-eval-client-latest.jar --systems http://localhost:8080/match --track <location-URI> <collection-name> <version>
+```