Impossible to train on CPU
n2oblife opened this issue · 1 comments
Hi !
When I try to train a customized pipeline on cpu the console returns an error, here is the core of the code :
`training_config={
'category': 'customized',
'task': 'posdep',
'save_dir': './save_dir',
'gpu' : False,
'train_conllu_fpath': my-path/train.conllu', # annotations file in CONLLU format for training
'dev_conllu_fpath': my-path/dev.conllu' # annotations file in CONLLU format for development
}
trainer = TPipeline(training_config)
trainer.train()'
and here is the output :
'File "/.../trankit_build/trankit/models/classifiers.py", line 130, in forward
diag = torch.eye(batch.head_idxs.size(-1) + 1, dtype=torch.bool).cuda().unsqueeze(0)
File "/.../lib/python3.10/site-packages/torch/cuda/init.py", line 247, in _lazy_init
torch._C._cuda_init()
RuntimeError: No CUDA GPUs are available'
I cloned the repo and changed the line to enable the training on CPU but wanted to warn you just in case (even if training on cpu is not efficient, some might not have the proper material, or like me want to test training on cpu before launching the scripts on dedicated servers)