Skip to content

Latest commit

 

History

History
43 lines (27 loc) · 1.67 KB

wmt14_en2fr.md

File metadata and controls

43 lines (27 loc) · 1.67 KB

Pre-processing, Training, and Evaluation on WMT'14 En-Fr dataset

This file describes the steps for (1) downloading dataset, (2) processing dataset, (3) training, and (4) evaluation.

Dataset download and pre-processing

From the main directory, run the following command:

bash prepare_nmt_dataset.sh wmt14_en_fr

Training

To train a model with a single node comprising of 8 V100 GPUs (each with 32 GB memory), you can use the following command:

python nmt_wmt14_en2fr.py --d-m 256

where --d-m is the model dimension. In our experiments, we have only tested d-m={128, 256, 384, 512, 640}

Evaluation

To evaluate a model, you can use the following command:

python generate.py data-bin/wmt14_en_fr/ --path <results_dir>/checkpoint_best.pt --beam 5 --lenpen 0.9 --remove-bpe --batch-size 128 --quiet

Results

Here are the results that we obtain.

Model dimension (d_m) Parameters BLEU Training Logs
128 8.19 M 34.1 Link
256 13.89 M 37.3 Link
384 23.35 M 38.3 Link
512 36.86 M 39.6 Link
640 54.14 M 40.5 Link