Open David-LiCause opened 4 years ago
When training the RetinaNet model on CPU or GPU with '--distribution_strategy=mirrored' an error occurs while exporting the trained model.
The error occurs here (https://github.com/tensorflow/tpu/blob/master/models/official/retinanet/retinanet_main.py#L704) when a tf.TPUEstimator instance is created with the 'config' set to an instance of tf.estimator.RunConfig instead of tf.contrib.tpu.RunConfig.
Error message:
INFO:tensorflow:Exporting saved model. I1022 20:36:52.486371 139714595177920 retinanet_main.py:695] Exporting saved model. Traceback (most recent call last): File "tpu/models/official/retinanet/retinanet_main.py", line 727, in <module> app.run(main) File "/usr/local/lib/python2.7/dist-packages/absl/app.py", line 299, in run _run_main(main, args) File "/usr/local/lib/python2.7/dist-packages/absl/app.py", line 250, in _run_main sys.exit(main(argv)) File "tpu/models/official/retinanet/retinanet_main.py", line 710, in main params=eval_params) File "/usr/local/lib/python2.7/dist-packages/tensorflow_estimator/python/estimator/tpu/tpu_estimator.py", line 2551, in __init__ '`config` must be provided with type `tpu_config.RunConfig`') ValueError: `config` must be provided with type `tpu_config.RunConfig`
Have you solved this problem? @David-LiCause
@Byronnar No, I haven't created a fix for this
When training the RetinaNet model on CPU or GPU with '--distribution_strategy=mirrored' an error occurs while exporting the trained model.
The error occurs here (https://github.com/tensorflow/tpu/blob/master/models/official/retinanet/retinanet_main.py#L704) when a tf.TPUEstimator instance is created with the 'config' set to an instance of tf.estimator.RunConfig instead of tf.contrib.tpu.RunConfig.
Error message: