Upload rl_code/examples/qwen2_5_vl_3b_geo3k_grpo.sh with huggingface_hub

rl_code/examples/qwen2_5_vl_3b_geo3k_grpo.sh

@@ -0,0 +1,16 @@
+#!/bin/bash
+
+set -x
+
+export PYTHONUNBUFFERED=1
+
+MODEL_PATH=Qwen/Qwen2.5-VL-3B-Instruct  # replace it with your local file path
+
+python3 -m verl.trainer.main \
+    config=examples/config.yaml \
+    data.train_files=hiyouga/geometry3k@train \
+    data.val_files=hiyouga/geometry3k@test \
+    worker.actor.model.model_path=${MODEL_PATH} \
+    worker.rollout.tensor_parallel_size=1 \
+    trainer.experiment_name=qwen2_5_vl_3b_geo_grpo \
+    trainer.n_gpus_per_node=2