NVIDIA-Merlin / NVIDIA-Merlin/Transformers4Rec
[FEA] Add support to ELECTRA discriminator
Open
@sararb is already working on this.
Since Oct 4, 2021.
area/transformers
P1
- Dominant language
- Python
- Stars
- 1.3k
- Forks
- 165
- Avg merge
- 1m
- Merged PRs (30d)
- 2
Description
🚀 Feature request
Define Generator-Discriminator architecture needed by ELECTRA task.
Motivation
ELECTRA model is a joint-training task using two Transformer blocks: The first block is trained with Masked Language Modeling, and the second task is trained with replacement token detection. The final loss is a weighted sum over these two tasks.
The logic of replacement token sampling can be found in the old version of the API here
Your contribution
Include the necessary tools for building, training, and evaluating the ELECTRA-based model.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Assessment
This issue has not been assessed yet.