update readme

- fix gt symbol appearing as quotes
- add HumanEval dataset in evaluation
This commit is contained in:
bentrevett 2021-07-17 16:28:45 +01:00
parent 95df5c7747
commit a8cd4aae55

View File

@ -15,8 +15,8 @@ GPT-Code-Clippy (GPT-CC) is an open source version of [GitHub Copilot](https://c
The dataset used to train GPT-CC is obtained from [SEART GitHub Search](https://seart-ghs.si.usi.ch/) using the following criteria:
- >10 GitHub stars
- >2 commits
- >10 GitHub stars
- >2 commits
- Must have a licence
- Exclude forks
- Size < 70708 bytes
@ -61,7 +61,7 @@ TODO: which is the recommended way to train GPT-CC?
## Evaluation
The models are also evaluated on the [APPS dataset](https://github.com/hendrycks/apps).
The models are also evaluated on the [APPS](https://github.com/hendrycks/apps) and [HumanEval](https://github.com/openai/human-eval) datasets.
TODO: evaluation results.