Model Prediction call stuck at convert_examples_to_features when using `use_multiprocessing=True`

Describe the bug Model prediction is failing while converting_examples_to_features when we use multiprocessing

To Reproduce Run a Gunicorn server which accepts prediction calls and hit predict method at classification_model.ClassificationModel().predict()

Expected behavior It should predict without any issues

Desktop (please complete the following information):

• Ubuntu 18.04

Additional context Observed replacing code at https://github.com/ThilinaRajapakse/simpletransformers/blob/master/simpletransformers/classification/classification_utils.py#L376 Fixes the issue

with Pool(process_count) as p:
               features = list(
                   tqdm(
                       p.imap(convert_example_to_feature, examples, chunksize=500),
                       total=len(examples),
                       disable=silent,
                   )
               )

to

    try:                                                                                                                                                                                                                                                                                
        p = Pool(process_count)                                                                                                                                                                                                                                                         
        features = list(                                                                                                                                                                                                                                                                    
            tqdm(                                                                                                                                                                                                                                                                               
                p.imap(convert_example_to_feature, examples, chunksize=500),                                                                                                                                                                                                                    
                            total=len(examples),                                                                                                                                                                                                                                                            
                            disable=silent,                                                                                                                                                                                                                                                             
            )                                                                                                                                                                                                                                                                           
       )                                                                                                                                                                                                                                                                           
    finally:                                                                                                                                                                                                                                                                            
        p.close()                                                                                                                                                                                                                                                                       
        p.join()

or

    with Pool(process_count) as p:                                                                                                                                                                                                                                                      
        features = list(                                                                                                                                                                                                                                                                    
            tqdm(                                                                                                                                                                                                                                                                               
                p.imap(convert_example_to_feature, examples, chunksize=500),                                                                                                                                                                                                                    
                            total=len(examples),                                                                                                                                                                                                                                                            
                            disable=silent,                                                                                                                                                                                                                                                             
                )                                                                                                                                                                                                                                                                           
            )                                                                                                                                                                                                                                                                               
        p.close()                                                                                                                                                                                                                                                                       
        p.join()

`

[Lysimachos]

I have the same problem. The simpletransformer library I am using is without the correction you are proposing. On what machine are you running the code on?

[nmvijay]

sometime just gets stuck (hangs), sometimes throws error (random behaviour)

0%| | 0/72685 [00:00<?, ?it/s] 0%| | 1/72685 [00:00<5:51:11, 3.45it/s] 2%|▏ | 1501/72685 [00:00<4:00:47, 4.93it/s] 4%|▍ | 3001/72685 [00:00<2:45:02, 7.04it/s] 6%|▌ | 4501/72685 [00:00<1:53:03, 10.05it/s] 8%|▊ | 6001/72685 [00:00<1:17:26, 14.35it/s] 10%|█ | 7501/72685 [00:00<53:00, 20.49it/s]
12%|█▏ | 9001/72685 [00:01<36:16, 29.25it/s] 14%|█▍ | 10012/72685 [00:01<25:06, 41.60it/s] 16%|█▌ | 11501/72685 [00:01<17:11, 59.34it/s] 18%|█▊ | 13001/72685 [00:01<11:45, 84.58it/s] 20%|█▉ | 14501/72685 [00:01<08:03, 120.40it/s] 22%|██▏ | 16001/72685 [00:01<05:31, 171.21it/s] 24%|██▍ | 17501/72685 [00:02<03:47, 243.03it/s] 26%|██▌ | 19001/72685 [00:02<02:36, 343.75it/s] 28%|██▊ | 20500/72685 [00:02<00:06, 8688.71it/s] multiprocessing.pool.RemoteTraceback: “”" Traceback (most recent call last): File “/usr/lib/python3.7/multiprocessing/pool.py”, line 121, in worker result = (True, func(*args, **kwds)) File “/usr/lib/python3.7/multiprocessing/pool.py”, line 44, in mapstar return list(map(*args)) File “/my_bundle_bundle/my_env/lib/python3.7/site-packages/simpletransformers/classification/classification_utils.py”, line 112, in convert_example_to_feature tokens_a = tokenizer.tokenize(example.text_a) File “/my_bundle_bundle/my_env/lib/python3.7/site-packages/transformers/tokenization_utils.py”, line 1329, in tokenize tokenized_text = split_on_tokens(added_tokens, text) File “/my_bundle_bundle/my_env/lib/python3.7/site-packages/transformers/tokenization_utils.py”, line 1303, in split_on_tokens if not text.strip(): AttributeError: ‘float’ object has no attribute ‘strip’ “”"

The above exception was the direct cause of the following exception:

Traceback (most recent call last): File “src/component_prediction_hf.py”, line 53, in <module> model.train_model(train_data) File “/my_bundle_bundle/my_env/lib/python3.7/site-packages/simpletransformers/classification/classification_model.py”, line 274, in train_model train_dataset = self.load_and_cache_examples(train_examples, verbose=verbose) File “/my_bundle_bundle/my_env/lib/python3.7/site-packages/simpletransformers/classification/classification_model.py”, line 852, in load_and_cache_examples args=args, File “/my_bundle_bundle/my_env/lib/python3.7/site-packages/simpletransformers/classification/classification_utils.py”, line 391, in convert_examples_to_features disable=silent, File “/my_bundle_bundle/my_env/lib/python3.7/site-packages/tqdm/std.py”, line 1129, in iter for obj in iterable: File “/usr/lib/python3.7/multiprocessing/pool.py”, line 325, in <genexpr> return (item for chunk in result for item in chunk) File “/usr/lib/python3.7/multiprocessing/pool.py”, line 748, in next raise value AttributeError: ‘float’ object has no attribute ‘strip’