torch.load() throws Segmentation fault (core dumped)

[BaiGeiQiShi]

Dear author, When I use python generator.py to generate patches, I got the Segmentation fault (core dumped). I found it was caused by torch.load().

def generate_gpt_fconv(vocab_file, model_file, input_file, identifier_txt_file, identifier_token_file, output_file, beam_size):
    dictionary = Dictionary(vocab_file, min_cnt=0)
    print(len(dictionary))
    loaded = torch.load(
            model_file, map_location="cpu"
    )

I tried map_location="cpu", map_location="cuda:0". My GPU is A5000, cuda==10.1, pytorch==1.4.0, python==3.8. The models I used are your pre-traned model.

Could you help me? Thanks in advance!

[BaiGeiQiShi]

Here I found the problem:

Exception has occurred: ModuleNotFoundError
No module named 'transformers.configuration_openai'
  File "/CURE/src/tester/generator.py", line 61, in generate_gpt_conut
    loaded = torch.load(
  File "/CURE/src/tester/generator.py", line 134, in <module>
    generate_gpt_conut(vocab_file, model_file, input_file, identifier_txt_file, identifier_token_file, output_file, beam_size)
ModuleNotFoundError: No module named 'transformers.configuration_openai'

But I already used pip to install transformers=2.10.0, is there a problem with my installation?