Tensorflow Troubleshooting

tfds.load always download even if data_dir is point to a local directory with pre-downloaded datasets

reason: maybe the local dataset is corrupted, try decompress it and open it to check it’s sanity.
solution: download it again!

reason: data_dir must point to a parent dir that contain a subdir named <dataset_name>

---

The transmission of request 0x7f8a6e3741e0 (URI: https://www.googleapis.com/storage/v1/b/tfds-data/o/dataset_info%2Fmnist%2F3.0.1?fields=size%2Cgeneration%2Cupdated) has been stuck at 0 of 0 bytes for 61 seconds and will be aborted. CURL timing information: lookup time: 0.00461 (No error), connect time: 2.90642 (No error), pre-transfer time: 0 (No error), start-transfer time: 0 (No error)

solution: if your vpn is ok, then just wait, as described by this issue.

---

java.net.URISyntaxException: Expected authority at index 7: hdfs://
solution:

---

tensorflow.python.framework.errors_impl.Unimplementederror: File system scheme hdfs not implemented.

reason: bug of tensorflow 1.2.1 on centos
solution: unintall tensorflow and reinstall it pip install tensorflow-1.2.1-cp35-cp35mu-manylinux1_x86_64.whl refer to this comment

---

Symbol not found: _H5Pget_fapl_ros

solution: As i was using conda, I just ran conda install h5py

---