RuntimeError: Found dtype Long but expected Float

[greed2411]

dependency versions:

simpletransformers (0.60.9)
torch (1.9.0)

when one tries to create a classification model with just one unique label, this happens. I'm assuming simpletransformers takes care of the tensor type differently when the no. of unique labels is 1 (and opp type, when it's greater than 1 -> usual scenario).

actual stack trace:

2021-07-28 22:24:07:851 slu [train.py:54] INFO Training started.
Epochs 0/1. Running Loss:    0.1531:   0%|                                                                                            | 0/2 [00:00<?, ?it/s]
Epoch 1 of 1:   0%|                                                                                                                   | 0/1 [00:00<?, ?it/s]
Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "/root/jaivarsan/to_be_deleted/yellow/slu/dev/cli.py", line 100, in main
    train_intent_classifier(config, file_format=file_format)
  File "/root/jaivarsan/to_be_deleted/yellow/slu/dev/train.py", line 55, in train_intent_classifier
    model.train_model(
  File "/root/.pyenv/versions/3.8.9/envs/dgblue/lib/python3.8/site-packages/simpletransformers/classification/classification_model.py", line 463, in train_model
    global_step, training_details = self.train(
  File "/root/.pyenv/versions/3.8.9/envs/dgblue/lib/python3.8/site-packages/simpletransformers/classification/classification_model.py", line 720, in train
    loss.backward()
  File "/root/.pyenv/versions/3.8.9/envs/dgblue/lib/python3.8/site-packages/torch/_tensor.py", line 255, in backward
    torch.autograd.backward(self, gradient, retain_graph, create_graph, inputs=inputs)
  File "/root/.pyenv/versions/3.8.9/envs/dgblue/lib/python3.8/site-packages/torch/autograd/__init__.py", line 147, in backward
    Variable._execution_engine.run_backward(
RuntimeError: Found dtype Long but expected Float

[greed2411]

supposedly every label should have (strictly) greater than 1 observations too. else one can face this error.

we require dataset validators from our end and, say the reason why it can fail (with warnings maybe).