Creating pull request of hacks I needed to run sample.py on cuda

[DavidLKing]

When I run even the basic sample.py script under examples on a cuda enabled machine, I still get errors that not all the vectors are cuda vectors (some are still cpu). You can ignore my edits in sample.py, but I did annotate in each place where I needed to add vector = vector.cuda() to make sample.py run. This occurs even when torch.device('cuda') is called, which should not be the case in PyTorch 0.4.0+.

Thank you, and feel free to follow up with any questions.

[pskrunner14]

Hi @DavidLKing did you try the develop branch? The necessary changes for moving tensors to CUDA have already been made there. Master branch is yet to be updated.

[DavidLKing]

Hi @pskrunner14! I did, but it still fails when I try to run sample.py with the same error:

(pytorch-seq2seq) david@Arjuna:~/bin/git/pytorch-seq2seq$ python examples/sample.py --train_path data/toy_reverse/train/data.txt --dev_path data/toy_reverse/dev/data.txt 
/home/david/miniconda2/envs/pytorch-seq2seq/lib/python3.6/site-packages/torch/nn/functional.py:52: UserWarning: size_average and reduce args will be deprecated, please use reduction='elementwise_mean' instead.
  warnings.warn(warning.format(ret))
2018-11-11 15:23:53,307 root         INFO     Namespace(dev_path='data/toy_reverse/dev/data.txt', expt_dir='./experiment', load_checkpoint=None, log_level='info', resume=False, train_path='data/toy_reverse/train/data.txt')
/home/david/miniconda2/envs/pytorch-seq2seq/lib/python3.6/site-packages/torch/nn/functional.py:52: UserWarning: size_average and reduce args will be deprecated, please use reduction='sum' instead.
  warnings.warn(warning.format(ret))
/home/david/miniconda2/envs/pytorch-seq2seq/lib/python3.6/site-packages/torch/nn/modules/rnn.py:38: UserWarning: dropout option adds dropout after all but last recurrent layer, so non-zero dropout expects num_layers greater than 1, but got dropout=0.2 and num_layers=1
  "num_layers={}".format(dropout, num_layers))
2018-11-11 15:23:56,082 seq2seq.trainer.supervised_trainer INFO     Optimizer: Adam (
Parameter Group 0
    amsgrad: False
    betas: (0.9, 0.999)
    eps: 1e-08
    lr: 0.001
    weight_decay: 0
), Scheduler: None
Traceback (most recent call last):
  File "examples/sample.py", line 129, in <module>
    resume=opt.resume)
  File "/home/david/miniconda2/envs/pytorch-seq2seq/lib/python3.6/site-packages/seq2seq-0.1.6-py3.6.egg/seq2seq/trainer/supervised_trainer.py", line 186, in train
  File "/home/david/miniconda2/envs/pytorch-seq2seq/lib/python3.6/site-packages/seq2seq-0.1.6-py3.6.egg/seq2seq/trainer/supervised_trainer.py", line 103, in _train_epoches
  File "/home/david/miniconda2/envs/pytorch-seq2seq/lib/python3.6/site-packages/seq2seq-0.1.6-py3.6.egg/seq2seq/trainer/supervised_trainer.py", line 55, in _train_batch
  File "/home/david/miniconda2/envs/pytorch-seq2seq/lib/python3.6/site-packages/torch/nn/modules/module.py", line 477, in __call__
    result = self.forward(*input, **kwargs)
  File "/home/david/miniconda2/envs/pytorch-seq2seq/lib/python3.6/site-packages/seq2seq-0.1.6-py3.6.egg/seq2seq/models/seq2seq.py", line 48, in forward
  File "/home/david/miniconda2/envs/pytorch-seq2seq/lib/python3.6/site-packages/torch/nn/modules/module.py", line 477, in __call__
    result = self.forward(*input, **kwargs)
  File "/home/david/miniconda2/envs/pytorch-seq2seq/lib/python3.6/site-packages/seq2seq-0.1.6-py3.6.egg/seq2seq/models/EncoderRNN.py", line 68, in forward
  File "/home/david/miniconda2/envs/pytorch-seq2seq/lib/python3.6/site-packages/torch/nn/modules/module.py", line 477, in __call__
    result = self.forward(*input, **kwargs)
  File "/home/david/miniconda2/envs/pytorch-seq2seq/lib/python3.6/site-packages/torch/nn/modules/sparse.py", line 110, in forward
    self.norm_type, self.scale_grad_by_freq, self.sparse)
  File "/home/david/miniconda2/envs/pytorch-seq2seq/lib/python3.6/site-packages/torch/nn/functional.py", line 1110, in embedding
    return torch.embedding(weight, input, padding_idx, scale_grad_by_freq, sparse)
RuntimeError: Expected object of type torch.cuda.LongTensor but found type torch.LongTensor for argument #3 'index'

I'm fine with this not being a full merge. Would an issue make more sense?

[pskrunner14]

Hey @DavidLKing are you sure you have the latest changes? It seems to be working fine on my side:

psk@ubuntu:~/projects/pytorch-seq2seq$ python examples/sample.py $TRAIN_SRC $TRAIN_TGT $DEV_SRC $DEV_TGT
2018-11-12 02:19:04,671:root:INFO: train_source: data/toy_reverse/train/src.txt
2018-11-12 02:19:04,671:root:INFO: train_target: data/toy_reverse/train/tgt.txt
2018-11-12 02:19:04,671:root:INFO: dev_source: data/toy_reverse/dev/src.txt
2018-11-12 02:19:04,671:root:INFO: dev_target: data/toy_reverse/dev/tgt.txt
2018-11-12 02:19:04,671:root:INFO: experiment_directory: ./experiment
2018-11-12 02:19:06,860:seq2seq.trainer.supervised_trainer:INFO: Optimizer: Adam (
Parameter Group 0
    amsgrad: False
    betas: (0.9, 0.999)
    eps: 1e-08
    initial_lr: 0.001
    lr: 0.001
    weight_decay: 0
), Scheduler: <torch.optim.lr_scheduler.StepLR object at 0x7fd4f1c03c88>
Train Perplexity: 8.6506: 100%|██████████████████████████████████████| 157/157 [00:03<00:00, 45.63it/s]
2018-11-12 02:19:10,454:seq2seq.trainer.supervised_trainer:INFO: Finished epoch 1: Train Perplexity: 6.4669, Dev Perplexity: 4.7727, Accuracy: 0.6091
Train Perplexity: 1.1459: 100%|██████████████████████████████████████| 157/157 [00:03<00:00, 47.59it/s]
2018-11-12 02:19:13,900:seq2seq.trainer.supervised_trainer:INFO: Finished epoch 2: Train Perplexity: 2.4010, Dev Perplexity: 4.4814, Accuracy: 0.7656
Train Perplexity: 1.0013: 100%|██████████████████████████████████████| 157/157 [00:03<00:00, 46.45it/s]
2018-11-12 02:19:17,427:seq2seq.trainer.supervised_trainer:INFO: Finished epoch 3: Train Perplexity: 1.2013, Dev Perplexity: 1.0010, Accuracy: 1.0000
['3', '2', '1', '<eos>']

[DavidLKing]

Okay, it seems there have been more drastic changes. I was using an older version of sample.py. The newer version runs fine with the develop branch. I'll need to update my own project, but I think we're good to go. Thanks for the time! When's the next release scheduled?

[pskrunner14]

Hi @DavidLKing yes there have been quite a few changes in the core API. Hopefully by next month it will be ready, still have a few bugs in the new features.