search

facebookresearch / XLM

PyTorch original implementation of Cross-lingual Language Model Pretraining.
Other
2.87k stars 495 forks source link

Error in Training #343

Open sadanyh opened 2 years ago

sadanyh commented 2 years ago

Hi,

I am getting this error while training. Is it a problem with APEX version? I am working on COLAB and have installed (apex-0.1). Could you please advise on this error? Thank you

INFO - 12/01/21 16:15:03 - 0:00:02 - Number of parameters (model): 168439772 INFO - 12/01/21 16:15:06 - 0:00:05 - Found 0 memories. INFO - 12/01/21 16:15:06 - 0:00:05 - Found 12 FFN. INFO - 12/01/21 16:15:06 - 0:00:05 - Found 198 parameters in model. Traceback (most recent call last): File "train.py", line 327, in main(params) File "train.py", line 240, in main trainer = SingleTrainer(model, data, params) File "/content/XLM/xlm/trainer.py", line 799, in init super().init(data, params) File "/content/XLM/xlm/trainer.py", line 58, in init assert params.amp >= 1 or not params.fp16 AssertionError

vicky-yuan commented 1 year ago

Hi, I had the same problem. Did you solve it? Looking forward to your reply!

sadanyh commented 1 year ago

Hi, I don't quite remember how I solved the issue. But I followed this notebook: https://colab.research.google.com/github/Tikquuss/meta_XLM/blob/master/notebooks/demo/tuto.ipynb This one worked fine.

vicky-yuan commented 1 year ago

Hi, I don't quite remember how I solved the issue. But I followed this notebook: https://colab.research.google.com/github/Tikquuss/meta_XLM/blob/master/notebooks/demo/tuto.ipynb This one worked fine.

Thank you very much, I solved the problem!