Error processing property '_dropout_mask_cache' of <ContextValueCache>

[gbaulard]

Error processing property '_dropout_mask_cache' of

from code2seq.

[lapplislazuli]

For other people coming here from the 2.1 Fork

The following requirements.txt worked for me:

clang
joblib
libclang
lmdb
networkx
numpy==1.19.5
rouge==0.3.2
tensorflow==2.1
tensorflow-gpu==2.1
keras==2.1
tensorflow_addons==0.9.1
scikit_learn
flatbuffers

However if your GPU is recognized depends on the model - newer GPUs do not support the old CUDA Frameworks required for TF 2.1.
3080Ti is not compatible.

[ethansaurusrex]

Hi @gbaulard,

Did you ever find a solution to this issue, I have run into the same problem in the official version and the tensor flow 2.1+ fork by Kolkir. Thanks!

[urialon]

Hi @gbaulard ,
Thank you for your interest in our work!

What python version and tensorflow version are you using?

Best,
Uri

[gbaulard]

Hi,

We tried :

• python 3.9.7 with tensorflow 2.8.0
• python 3.8.10 with tensorflow 2.8.0
  with the unofficial version of the project.

With your official version, it was python 3.6.5 and tensorflow 1.12.0

The 3 configs give this same error, running train.sh bash WSL or code2seq.py in conda.

Best

[urialon]

Hi all,
I think that the code supports only tensorflow 1.*.

If you are interested in NL<->code, I recommend checking our newer projects which are all publicly available using the Huggingface library:
https://github.com/VHellendoorn/Code-LMs
https://github.com/neulab/code-bert-score
https://github.com/shuyanzhou/docprompting

Best,
Uri

[ethansaurusrex]

Hi,

Thank you for the quick reply! I will definitely take a look at those papers.

Best