Benchmark Submission: 0bserver07

[0bserver07]

This pull request represents a submission to the codesearchnet benchmark.

It is the main communication channel for you and the reviewers.

Wandb run results for review

[hamelsmu]

@0bserver07 can you tell us a little about your benchmark submission and how you produced the results? What did you change exactly etc, your thought process etc? Please share!

[0bserver07]

hey @hamelsmu, absolutely.

I changed the params to the following:

train.py —max-num-epochs 50 —test-batch-size 250 —run-name baseline-expv001

I was planning to reduce the batch size and make the model to wider. With the following configs. But it turns out, the parameters I tried passing had a space in them, and it didn't pass down to the model's function.

The reasoning between the above and the current experiment I'm running is for reproducing a toy benchmark that I had for comparing grammar of text and program AST Trees.

It originated from a dataset that can from this research question here.

---

ps: I wasn't sure if I should merge the change as a branch on top of this repository or separately allowed to submit results.

[hamelsmu]

@0bserver07 thanks for the explanation! That is helpful, merging now. cc: @mallamanis @hohsiangwu