Error on gpus set to 0

[arsalan993]

Although i set gpus = 0 since i dont have any GPU installed scripted did predicted the output and saved it in prediction.txt file but right before exiting run.py file compiler throughs this error

Starting re-scoring ...

{'<arg1>': '[unused1]', '</arg1>': '[unused2]', '<rel>': '[unused3]', '</rel>': '[unused4]', '<arg2>': '[unused5]', '</arg2>': '[unused6]', 'SENT': '[unused7]', 'PRED': '[unused8]', '@COPY@': '[unused9]', 'EOE': '[unused10]'}
Traceback (most recent call last):
  File "run.py", line 469, in <module>
    main(hyperparams)
  File "run.py", line 459, in main
    train_dataloader, val_dataloader, test_dataloader, all_sentences)
  File "run.py", line 255, in splitpredict
    rescored = rescore(inp_fp, model_dir=hparams.rescore_model, batch_size=256)
  File "imojie/imojie/aggregate/score.py", line 90, in rescore
    return generate_probs(model_dir, inp_fp, weights_fp, topk, out_ext, cuda_device, overwrite=overwrite, extraction_ratio=ext_ratio, batch_size=batch_size, out=None)
  File "imojie/imojie/aggregate/score.py", line 39, in generate_probs
    probs = evaluate_from_args(args)
  File "/home/arsalan/Documents/openie6/topic/lib/python3.6/site-packages/allennlp/commands/evaluate.py", line 131, in evaluate_from_args
    archive = load_archive(args.archive_file, args.cuda_device, args.overrides, args.weights_file)
  File "/home/arsalan/Documents/openie6/topic/lib/python3.6/site-packages/allennlp/models/archival.py", line 230, in load_archive
    cuda_device=cuda_device)
  File "/home/arsalan/Documents/openie6/topic/lib/python3.6/site-packages/allennlp/models/model.py", line 329, in load
    return cls.by_name(model_type)._load(config, serialization_dir, weights_file, cuda_device)
  File "/home/arsalan/Documents/openie6/topic/lib/python3.6/site-packages/allennlp/models/model.py", line 277, in _load
    model_state = torch.load(weights_file, map_location=util.device_mapping(cuda_device))
  File "/home/arsalan/Documents/openie6/topic/lib/python3.6/site-packages/torch/serialization.py", line 386, in load
    return _load(f, map_location, pickle_module, **pickle_load_args)
  File "/home/arsalan/Documents/openie6/topic/lib/python3.6/site-packages/torch/serialization.py", line 573, in _load
    result = unpickler.load()
  File "/home/arsalan/Documents/openie6/topic/lib/python3.6/site-packages/torch/serialization.py", line 536, in persistent_load
    deserialized_objects[root_key] = restore_location(obj, location)
  File "/home/arsalan/Documents/openie6/topic/lib/python3.6/site-packages/torch/serialization.py", line 409, in restore_location
    result = map_location(storage, location)
  File "/home/arsalan/Documents/openie6/topic/lib/python3.6/site-packages/allennlp/nn/util.py", line 828, in inner_device_mapping
    return storage.cuda(cuda_device)
  File "/home/arsalan/Documents/openie6/topic/lib/python3.6/site-packages/torch/_utils.py", line 69, in _cuda
    with torch.cuda.device(device):
  File "/home/arsalan/Documents/openie6/topic/lib/python3.6/site-packages/torch/cuda/__init__.py", line 243, in __enter__
    self.prev_idx = torch._C._cuda_getDevice()
  File "/home/arsalan/Documents/openie6/topic/lib/python3.6/site-packages/torch/cuda/__init__.py", line 178, in _lazy_init
    _check_driver()
  File "/home/arsalan/Documents/openie6/topic/lib/python3.6/site-packages/torch/cuda/__init__.py", line 99, in _check_driver
    http://www.nvidia.com/Download/index.aspx""")
AssertionError: 
Found no NVIDIA driver on your system. Please check that you
have an NVIDIA GPU and installed a driver from
http://www.nvidia.com/Download/index.aspx

If you suspect this is an IPython 7.16.1 bug, please report it at:
    https://github.com/ipython/ipython/issues
or send an email to the mailing list at [email protected]

You can print a more detailed traceback right now with "%tb", or use "%debug"
to interactively debug it.

Extra-detailed tracebacks for bug-reporting purposes can be enabled via:
    %config Application.verbose_crash=True

[Martin36]

I get the same error. I have a GPU but I'm running Ubuntu in Windows so I can't use it without installing the Windows Insider beta version of Windows.

If a GPU is required to run, it would be nice to have that stated in the documentation.

But from what I can see, it seems that predictions on the input text were made. So maybe the error does not affect the results.

[sadakmed]

though it doesn't effect ur desired output bcoz it's generated before the error. However, copying the command from readme file as it is, including the argument --mode as splitpredict is the issue here.
setting it as predict will just predict and stop before the error. but the results will be not good as the first one.

[lapp0]

The issue lies in openie6/run.py.

rescored = rescore(inp_fp, model_dir=hparams.rescore_model, batch_size=256)

needs to be changed to

rescored = rescore(inp_fp, model_dir=hparams.rescore_model, batch_size=256, cuda_device=(0 if has_cuda else -1))

https://github.com/dair-iitd/openie6/blob/master/imojie/imojie/aggregate/score.py#L88 is called with cuda_device=0

This results in a call to

https://github.com/dair-iitd/openie6/blob/master/imojie/allennlp/allennlp/commands/evaluate.py#L92

with cuda device of 0, but as the documentation specifies, -1 is required for no GPU.

RUN sed -i 's|rescored = rescore(inp_fp, model_dir=hparams.rescore_model, batch_size=256)|rescored = rescore(inp_fp, model_dir=hparams.rescore_model, batch_size=256, cuda_device=(0 if has_cuda else -1))|' openie6/run.py worked in my dockerfile. A bit busy to make a PR, but may tackle it later.