CODEWITHSUNDEEP
condapermission-denied
liuhuo wan
liuhuo wan

Reputation: 21

Tensorflow failed because it has no write permission in conda environment

In the conda environment which was created by me, I try to run tensorflow-hub, it fails when tensorflow try to get write access.

error message:

tensorflow.python.framework.errors_impl.PermissionDeniedError:
 /tmp/tfhub_modules/96e8f1d3d4d90ce86b2db128249eb8143a91db73.lock.tmp40b1daf1ca7d4f62a23d9457654a995d;
Permission denied

error code:

module_url =
"https://tfhub.dev/google/universal-sentence-encoder-large/3"  embed =
hub.Module(module_url)

I know the reason is my py file has no write access, but I don't know how to solve it if I continue want to work in my created conda environment rather than my local machine.

whole error message :

File
"/home/ld-sgdev/liuhuo_wan/.conda/envs/tensorflow-hub/lib/python3.7/site-packages/tensorflow_hub/resolver.py",
line 379, in atomic_download
     overwrite=False)   File "/home/ld-sgdev/liuhuo_wan/.conda/envs/tensorflow-hub/lib/python3.7/site-packages/tensorflow_hub/tf_utils.py",
 line 65, in atomic_write_string_to_file
     f.write(contents)   File "/home/ld-sgdev/liuhuo_wan/.conda/envs/tensorflow-hub/lib/python3.7/site-packages/tensorflow/python/lib/io/file_io.py", line 106, in write
     self._prewrite_check()   File "/home/ld-sgdev/liuhuo_wan/.conda/envs/tensorflow-hub/lib/python3.7/site-packages/tensorflow/python/lib/io/file_io.py", line 92, in _prewrite_check
     compat.as_bytes(self.__name), compat.as_bytes(self.__mode)) tensorflow.python.framework.errors_impl.PermissionDeniedError:
 /tmp/tfhub_modules/96e8f1d3d4d90ce86b2db128249eb8143a91db73.lock.tmp40b1daf1ca7d4f62a23d9457654a995d;
 Permission denied

Upvotes: 2

Views: 1860

Answers (1)

Fei
Fei

Reputation: 1071

Encountered the same problem. In my case the following command solves it.

$ export TFHUB_CACHE_DIR=./tmp

I think the root cause is, when you use tf_hub to load something, it uses the default folder (/tmp) to save the downloaded model. In my case, my Tensorflow doesn't have the write access to the default folder.

Upvotes: 5

Related Questions