My jobs in gcloud ml-engine crash because the train file doesn't exist. In the screenshot you can see the error in the top line. At the bottom I check the contents of the directory and the file is clearly there.
Asked
Active
Viewed 247 times
1
Zachary Newman
- 20,014
- 4
- 39
- 37
Brendan von hofe
- 23
- 3
-
I don't think pandas supports reading files directly from gcloud, that's the problem. – Dr. Snoopy Feb 01 '18 at 08:04
1 Answers
1
Pandas does not have native GCS support. If the pandas method you are using accepts file object, then you can use the TensorFlow file_io library to open the file. See the detailed answer here. If not, then you can copy the file to VM first using gsutil cli.
Guoqing Xu
- 482
- 3
- 9