top-k predictions not generated

[nashid]

I trained the model with:

python -m nmt.nmt \
    --attention=scaled_luong \
    --src=vi --tgt=en \
    --vocab_prefix=/tmp/data/token  \
    --train_prefix=/tmp/data/train \
    --dev_prefix=/tmp/data/valid  \
    --test_prefix=/tmp/data/test \
    --out_dir=/tmp/model-top-k \
    --num_train_steps=10000 \
    --steps_per_stats=100 \
    --num_layers=2 \
    --num_units=1024 \
    --dropout=0.2 \
    --metrics=bleu \
    --optimizer=sgd \
    --learning_rate=1.0 \
    --start_decay_step=5000 \
    --decay_steps=10 \
    --encoder_type=bi \
    --beam_width=10

Followed to that used the inference engine with the following:

python -m nmt.nmt \
    --out_dir=./model \
    --inference_input_file=./data/test.buggy.beam.search \
    --inference_output_file=./data/testing-beam-search/model.output \
    --beam_width=10 \
    --num_translations_per_input=1 

I see in the log:

decoder: infer_mode=greedybeam_width=10, length_penalty=0.000000, coverage_penalty=0.000000
.....
.....
.....
  dynamic_seq2seq/decoder/attention/multi_rnn_cell/cell_3/basic_lstm_cell/bias:0, (512,), /device:GPU:0
  dynamic_seq2seq/decoder/attention/luong_attention/attention_g:0, (), /device:GPU:0
  dynamic_seq2seq/decoder/attention/attention_layer/kernel:0, (256, 128), /device:GPU:0
  dynamic_seq2seq/decoder/output_projection/kernel:0, (128, 10003), 
  loaded infer model parameters from ./code_model/translate.ckpt-20000, time 0.29s
# Start decoding
  decoding to output ./data/testing-beam-search/model.output
  done, num sentences 1, num translations per input 1, time 0s, Thu Jun 11 21:20:29 2020.

However, only one prediction is generated. Any ideas?