Fail to running on multi-gpus

inspurasc commented 6 years ago

ENV: os:ubuntu16.4 python: 2.7.5 tensorflow:1.8. cuda:9.2

While training WMT German-English, I request 4 gpus, but it running on 1 gpu. The cmd is below:

CUDA_VISIBLE_DEVICES=0,1,2,3 python -m nmt.nmt --src=de --tgt=en --hparams_path=nmt/standard_hparams/wmt16_gnmt_4_layer.json --out_dir=/tmp/deen_gnmt --vocab_prefix=/data/xinyx/wmt_data/vocab.bpe.32000 --train_prefix=/data/xinyx/wmt_data/train.tok.clean.bpe.32000 --dev_prefix=/data/xinyx/wmt_data/newstest2013.tok.bpe.32000 --test_prefix=/data/xinyx/wmt_data/newstest2015.tok.bpe.32000 --num_gpus=4 --num_train_steps=12000 --steps_per_stats=100 --num_layers=4 -num_units=128 --dropout=0.2 --metrics=bleu

Please give me some advice how i should do so that it can run on multiple GPU. bug

inspurasc commented 6 years ago

I have solve this problem.Thanks.

mohammedayub44 commented 5 years ago

@inspurasc How did you solve this problem if you can please post the solution.

Mohammed Ayub

swinalmeida commented 3 years ago

@inspurasc How did you resolve this issue. Can you please help!!

tensorflow / nmt

Fail to running on multi-gpus #346