Failed to generate create self-training dataset per transcoder-st doc

[weidotwisc]

Hi,

When I was following the instructions from https://github.com/facebookresearch/CodeGen/blob/main/docs/TransCoder-ST.md
to create self-training dataset. The create_self_training_dataset.sh failed with the Assertion Error:

######### Creating Tests ##########
Traceback (most recent call last):
File "codegen_sources/test_generation/create_tests.py", line 260, in
assert input_path.exists(), f"{input_path} does not exist"
AssertionError: /checkpoint/broz/data/2021-04-19_selected_sa_java_functions_for_tests_deduped does not exist

I am wondering how should I fix it ?

Thanks!

Wei

[brozi]

Hi,
Yes it should have been something that works on any cluster. Sorry about that.
This should fix it: 14a2f99