Hey, I’m trying to generate output of my model with alignements, so I use -report_align but I encounter this error:
[2023-07-06 15:03:47,001 INFO] Loading checkpoint from models/model.fren_step_52000.pt
[2023-07-06 15:03:47,619 INFO] Loading data into the model
['0-0.34585', '1-0.63270', '2-0.61036', '3-0.32734', '4-0.46401', '5-0.42423', '6-0.41460', '7-0.62102', '8-0.47739', '9-0.44198', '10-0.36853', '11-0.41361', '12-0.50141', '13-0.55924', '14-0.54243', '15-0.33698']
Traceback (most recent call last):
File "/home/jourdelune/Bureau/dev/translation-engine/venv/bin/onmt_translate", line 8, in <module>
sys.exit(main())
^^^^^^
File "/home/jourdelune/Bureau/dev/translation-engine/venv/lib/python3.11/site-packages/onmt/bin/translate.py", line 56, in main
translate(opt)
File "/home/jourdelune/Bureau/dev/translation-engine/venv/lib/python3.11/site-packages/onmt/bin/translate.py", line 37, in translate
_, _ = translator._translate(
^^^^^^^^^^^^^^^^^^^^^^
File "/home/jourdelune/Bureau/dev/translation-engine/venv/lib/python3.11/site-packages/onmt/translate/translator.py", line 421, in _translate
n_best_preds_align = [" ".join(align) for align in align_pharaohs]
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/jourdelune/Bureau/dev/translation-engine/venv/lib/python3.11/site-packages/onmt/translate/translator.py", line 421, in <listcomp>
n_best_preds_align = [" ".join(align) for align in align_pharaohs]
^^^^^^^^^^^^^^^
TypeError: sequence item 0: expected str instance, list found
I use this command: nmt_translate -model models/model.fren_step_52000.pt -src temps/.test.txt -output temps/.output.txt -report_align
Here is my model conf:
save_data: opennmt
# Training files
data:
corpus_1:
path_src: temps/src-train.subword
path_tgt: temps/tgt-train.subword
valid:
path_src: temps/src-val.subword
path_tgt: temps/tgt-val.subword
# Vocabulary files, generated by onmt_build_vocab
src_vocab: opennmt/source.vocab
tgt_vocab: opennmt/target.vocab
# Vocabulary size - should be the same as in sentence piece
src_vocab_size: 50000
tgt_vocab_size: 50000
# Tokenization options
src_subword_model: models/source.model
tgt_subword_model: models/target.model
# Where to save the log file and the output models/checkpoints
log_file: temps/train.log
save_model: models/model.fren
save_checkpoint_steps: 500
train_steps: 3000
valid_steps: 1000
early_stopping: 4
# Train on a single GPU
world_size: 1
gpu_ranks: [0]
keep_checkpoint: 3
seed: 3435
# Default: 4000 - for large datasets, try up to 8000
warmup_steps: 8000
report_every: 100
# Batching
bucket_size: 262144
batch_type: "tokens"
batch_size: 4096
valid_batch_size: 2048
max_generator_batches: 2
accum_count: [4]
accum_steps: [0]
# Optimization
model_dtype: "fp16"
optim: "adam"
learning_rate: 2
# warmup_steps: 8000
decay_method: "noam"
adam_beta2: 0.998
max_grad_norm: 0
label_smoothing: 0.1
param_init: 0
param_init_glorot: true
normalization: "tokens"
# Model
encoder_type: transformer
decoder_type: transformer
position_encoding: true
enc_layers: 6
dec_layers: 6
heads: 8
hidden_size: 512
word_vec_size: 512
transformer_ff: 2048
dropout_steps: [0]
dropout: [0.1]
attention_dropout: [0.1]
(without this argument, onmt_translate work fine)
.