YovaKem
YovaKem
Is the generator counterpart to this model: dbmdz/electra-small-turkish-cased-discriminator available? Thanks!
The OSCAR paper reports a CIDEr score of 78.8 and 80.9 for OSCAR base and large, resp. Since it's not clarified, I assume these are scores on the NoCaps test...
I trained an ELECTRA-small model following exactly the instructions in the README and obtained a model with downstream performance substantially worse than the google/electra-small-discriminator on HuggingFace. Apart from the training...
The README says ELECTRA-small has 256 hidden units and ELECTRA-base has 768. [Here](https://github.com/google-research/electra/blob/master/configure_pretraining.py) the embedding size for a small model is being set to 128 and for a base model...
In the paper you say " On the other hand, tying all encoder weights caused little improvement while incurring the significant disadvantage of requiring the generator and discriminator to be...
"In contrast to the original paper, the generated trees are always fitted with the same maximum depth. In the original implementation the maximum depth of the tree are drawn from...
Can you tell me what hyperparameters were used for the beam search at inference time and anything concerning penalty for length and repetition? Thanks!
Running the evaluation in the following fashion: ``` ./generate.py --model jkulhanek/augpt-mw-21 --dataset multiwoz-2.1-test --file predictions.txt ./evaluate_multiwoz.py --file predictions.txt --dataset multiwoz-2.1-test ``` I get inform and success results somewhat lower than...
Perhaps the data version should not be hardcoded [here](https://github.com/ufal/augpt/blob/fa8a57961ed1d8fe6099978c489c0b0f8956d64e/train_multiwoz.py#L43) in case one is training on MultiWOZ 2.0.