search

NVIDIA / OpenSeq2Seq

Toolkit for efficient experimentation with Speech Recognition, Text2Speech and NLP
https://nvidia.github.io/OpenSeq2Seq
Apache License 2.0
1.54k stars 369 forks source link

Inference using pre-trained Transformer #332

Closed sundeepteki closed 5 years ago

sundeepteki commented 5 years ago

Hi,

I downloaded the Transformer checkpoint and would like to evaluate its performance for EN-DE translation as per the command below: python3 run.py --config_file=example_configs/text2text/en-de/transformer-big.py --mode=infer --infer_output_file=raw.txt

It seems I'm missing a few files like the BPE tokens - wmt14-en-de.src.BPE_common.32K.tok wmt14-en-de.ref.BPE_common.32K.tok

Is it possible to upload these and other files required for EN-DE inference, as the scripts/get_big_en-de_data.sh takes a long time to process?

Thanks!

borisgin commented 5 years ago

No. These files are too big.