diff options
Diffstat (limited to 'tensorflow/docs_src/programmers_guide/using_tpu.md')
| -rw-r--r-- | tensorflow/docs_src/programmers_guide/using_tpu.md | 7 |
1 files changed, 4 insertions, 3 deletions
diff --git a/tensorflow/docs_src/programmers_guide/using_tpu.md b/tensorflow/docs_src/programmers_guide/using_tpu.md index a9c2cb3e33..d74d7f3181 100644 --- a/tensorflow/docs_src/programmers_guide/using_tpu.md +++ b/tensorflow/docs_src/programmers_guide/using_tpu.md @@ -129,9 +129,10 @@ my_tpu_estimator = tf.contrib.tpu.TPUEstimator( Typically the `FLAGS` would be set by command line arguments. To switch from training locally to training on a cloud TPU you would need to: -* Set `FLAGS.use_tpu` to `True` -* Set `FLAGS.tpu_name` so the `tf.contrib.cluster_resolver.TPUClusterResolver` can find it -* Set `FLAGS.model_dir` to a Google Cloud Storage bucket url (`gs://`). + 1) Set `FLAGS.use_tpu` to `True` + 1) Set `FLAGS.tpu_name` so the + `tf.contrib.cluster_resolver.TPUClusterResolver` can find it + 1) Set `FLAGS.model_dir` to a Google Cloud Storage bucket url (`gs://`). ## Optimizer |