try adding gpus=all option to debug gpu failures

Author: nitins17

.github/workflows/bazel_gpu_non_rbe.yml

@@ -42,7 +42,9 @@ jobs:
                         {name: "nosla", image: "us-central1-docker.pkg.dev/tensorflow-sigs/tensorflow/nosla-cuda12.3-cudnn9.1-ubuntu20.04-manylinux2014-multipython:latest"},
                         {name: "ml-build", image: "us-central1-docker.pkg.dev/tensorflow-sigs/tensorflow/ml-build:latest"}
                         ]
-    container: ${{ matrix.container-list.image }}
+    container: 
+      image: ${{ matrix.container-list.image }}
+      options: --gpus all
     env:
       JAXCI_HERMETIC_PYTHON_VERSION: 3.11
       JAXCI_ENABLE_X64: ${{ matrix.enable-x_64 }}