Methodology questions
#2
by
justinbarton
- opened
Hi @qilowoq . Could you comment on how you've ported this over from the published version of the model from the original authors?
For example, are you using the published trained weights, or have you pre-trained this version from scratch?
I looked at arcitecture and it was roberta but names of layers were different. It is published trained weights from open github repo
And aslo original model had custom embedding layer. U can see that in files. Its not a standard huggingface roberta