Update README.md

README.md

@@ -44,10 +44,29 @@ configs:
 [![Project Page](https://img.shields.io/badge/Website-000000?style=for-the-badge&logo=google-chrome&logoColor=white)]()
 </div>
 
-Here are the RL Data used to train **[OpenMMReasoner-RL](https://huggingface.co/OpenMMReasoner/OpenMMReasoner-RL)**. We use **[lmms-engine](https://github.com/EvolvingLMMs-Lab/lmms-engine)** as the training framework.
+Here are the RL Data used to train **[OpenMMReasoner-RL](https://huggingface.co/OpenMMReasoner/OpenMMReasoner-RL)**. We use **[verl](https://github.com/volcengine/verl)** as the training framework.
 
-To use this dataset, first snapshot-download the entire repository to your local machine. Then you can use the example script in the GitHub repository with your local data folder to use the parquet file.
+To use this dataset, first snapshot-download the entire repository to your local machine. After that, you can load the dataset using the example script provided in our GitHub repository by pointing it to your local data folder and the parquet file.
 
+An example configuration file would be:
+
+```bash
+DATA_FOLDER=/path/to/your/data
+
+ray job submit --address="http://127.0.0.1:8265" \
+    --runtime-env=verl/trainer/runtime_env.yaml \
+    -- \
+    bash -c "cd /path/to/your/verl/ && \
+    python3 -m verl.trainer.main_ppo \
+        algorithm.adv_estimator=${adv_estimator} \
+        actor_rollout_ref.actor.policy_loss.loss_mode=${loss_mode} \
+        data.train_files=[$DATA_FOLDER/algopuzzle_train.parquet,$DATA_FOLDER/mmk12_train.parquet,$DATA_FOLDER/puzzlevqa_train.parquet,$DATA_FOLDER/thinklite_vl_hard_train.parquet,$DATA_FOLDER/tqa_train.parquet,$DATA_FOLDER/virl39k_train.parquet,$DATA_FOLDER/wemath_standard.parquet,$DATA_FOLDER/wemath_pro.parquet] \
+        data.val_files=${DATA_FOLDER}/val.parquet \
+
+
+    ... rest of the command args ...
+
+```