keras_nmt

This projects provides an nmt implementation with keras. Data parallel training also supported. To run NMT experiments:

python train.py --state config.py > theano_nmt.log 2>&1 &

Name		Name	Last commit message	Last commit date
Latest commit History 50 Commits
data		data
eai_arkseg		eai_arkseg
lib		lib
.gitignore		.gitignore
README.md		README.md
algorithm.py		algorithm.py
aligner.py		aligner.py
backend.py		backend.py
config.py		config.py
configurations.py		configurations.py
data_stream.py		data_stream.py
datasets.py		datasets.py
detokenizer.perl		detokenizer.perl
initialization.py		initialization.py
mkl_gru.py		mkl_gru.py
mkl_gru_backward.py		mkl_gru_backward.py
mkl_gru_gradients.py		mkl_gru_gradients.py
mkl_gru_modified.py		mkl_gru_modified.py
mlsl_api.py		mlsl_api.py
models.py		models.py
multinode_cmd.sh		multinode_cmd.sh
nmt.py		nmt.py
pinyin.py		pinyin.py
prepare_training_data.py		prepare_training_data.py
preprocess.py		preprocess.py
print_val.py		print_val.py
reconstructor.py		reconstructor.py
replace_unk.py		replace_unk.py
run_nmt.sh		run_nmt.sh
sample_server.py		sample_server.py
sampling.py		sampling.py
sampling_cpus.py		sampling_cpus.py
search.py		search.py
sum_backward.py		sum_backward.py
sum_op.py		sum_op.py
test.py		test.py
test_multiple_gpu.py		test_multiple_gpu.py
tokenizer.perl		tokenizer.perl
train.py		train.py
train_on_multiple_hosts.py		train_on_multiple_hosts.py
train_on_multiple_hosts_sync.py		train_on_multiple_hosts_sync.py
utils.py		utils.py
worker.py		worker.py

Provide feedback