Open surajkota opened 4 years ago
Right - the dataclasses in the descriptor.py referenced in the stack trace need to be updated to include the [ml.script] section. This should also be followed up with possible changes in the transpiling done in the executor to ensure that the script is being set appropriately in the underlying k8s resource it is producing.
We should also add test coverage for this as it is a regression. It got introduced when we refactored the descriptor into dataclasses...
TOML file used: ec2_tf_cpu_single_node_synthetic.toml.txt
client keeps polling for status of the benchmark and does not recieve the error message hence marking it as related to #1001 #996
Error in executor:
recording some more part of log incase needed
client behavior: