File size: 682 Bytes
12d8f44 75c31f0 12d8f44 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 |
experiment_name: 'runs/transformer_huge'
dataset:
src_lang: 'lo'
src_tokenizer: 'BPE'
src_max_seq_len: 400
tgt_lang: 'vi'
tgt_tokenizer: 'WordLevel'
tgt_max_seq_len: 350
train_dataset: 'train_clean.dat'
validate_dataset: 'dev_clean.dat'
tokenizer_file: "tokenizer_{0}.json"
bleu_dataset: 'test2023'
model:
d_model: 512
num_heads: 8
d_ff: 2048
dropout_p: 0.15
num_encoder_layers: 8
num_decoder_layers: 4
model_folder: "weights"
model_basename: "transformer_huge"
preload: '_61'
train:
lr: 0.001 # 1e-2
batch_size: 32
num_epochs: 61
label_smoothing: 0.1
on_colab: False # are you training on Colab?
patience: 1
warm_up_steps: 200 |