nthngdy commited on
Commit
72aaf32
1 Parent(s): d451554

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +1 -1
README.md CHANGED
@@ -5,7 +5,7 @@ thumbnail: https://huggingface.co/front/thumbnails/google.png
5
  license: apache-2.0
6
  ---
7
 
8
- **WARNING**: This model is not scaled properly for pre-training with [google/electra-small-discriminator](https://huggingface.co/google/electra-small-discriminator). The ratio of numbers of parameters is 1:1 instead of 1:4. Pre-training using this config off the shelf will result in training instability and collapse of the discriminator loss.
9
 
10
  ## ELECTRA: Pre-training Text Encoders as Discriminators Rather Than Generators
11
 
 
5
  license: apache-2.0
6
  ---
7
 
8
+ **WARNING**: This is the official generator checkpoint as in the [ELECTRA original codebase](https://github.com/google-research/electra). However, this model is not scaled properly for pre-training with [google/electra-small-discriminator](https://huggingface.co/google/electra-small-discriminator). The paper recommends a hyperparameter multiplier of 1/4 between the discriminator and generator for this given model. This would not be the case when using `google/electra-small-generator` and `google/electra-small-discriminator`, which are similar in size.
9
 
10
  ## ELECTRA: Pre-training Text Encoders as Discriminators Rather Than Generators
11