Branchformer

MT config: conf/tuning/train_mt_branchformer_lr3e-3_warmup10k_share_enc_dec_input_dropout0.3.yaml
#Params: 43.40 M
Model link: https://huggingface.co/pyf98/iwslt14_de_en_branchformer

Environments

date: Sun May 29 01:39:59 EDT 2022
python version: 3.9.12 (main, Apr 5 2022, 06:56:58) [GCC 7.5.0]
espnet version: espnet 202205
pytorch version: pytorch 1.11.0
Git hash: 1cab3306f8136e614339390f59f06e11d054bbd8
- Commit date: Sat May 28 20:15:14 2022 -0400

dataset	score	verbose_score
beam5_maxlenratio1.6_penalty0.4/valid	34.1	67.6/42.9/29.0/20.0 (BP = 0.945 ratio = 0.946 hyp_len = 137200 ref_len = 144976)
beam5_maxlenratio1.6_penalty0.4/test	32.7	66.9/41.5/27.5/18.7 (BP = 0.945 ratio = 0.946 hyp_len = 121209 ref_len = 128122)

dataset	score	verbose_score
beam5_maxlenratio1.6_penalty0.4/valid	34.96	68.2/43.5/29.6/20.6 (BP=0.953, ratio=0.954, hyp_len=141414, ref_len=148225)
beam5_maxlenratio1.6_penalty0.4/test	33.72	67.6/42.3/28.2/19.3 (BP=0.954, ratio=0.955, hyp_len=125309, ref_len=131156)

mt_config: conf/tuning/train_mt_transformer_lr3e-3_warmup10k_share_enc_dec_input_dropout0.3.yaml
inference_config: conf/decode_mt.yaml

Metric: BLEU-4, detokenized case-sensitive BLEU result (single-reference)

dataset	bleu_score	verbose_score
beam5_maxlenratio1.6_penalty0.2/valid	33.3	68.4/42.9/28.9/19.8 (BP = 0.924 ratio = 0.927 hyp_len = 134328 ref_len = 144976)
beam5_maxlenratio1.6_penalty0.2/test	32.2	67.2/41.4/27.4/18.5 (BP = 0.933 ratio = 0.935 hyp_len = 119813 ref_len = 128122)

dataset	bleu_score	verbose_score
beam5_maxlenratio1.6_penalty0.2/valid	34.10	69.3/43.8/29.7/20.5 (BP=0.925, ratio=0.928, hyp_len=137509, ref_len=148225)
beam5_maxlenratio1.6_penalty0.2/test	33.13	68.3/42.4/28.3/19.3 (BP=0.934, ratio=0.936, hyp_len=122809, ref_len=131156)