Custom attention masks error running Trainer
Hello.
I have the following error while running Trainer to fine-tune a Decider model on my own data:
I've also taken the suggested steps and added naive_attention_prefill to True in config. But then get an error when loading the model:
ValueError: The model class you are passing has a `config_class` attribute that is not consistent with the config class you passed (model has and you passed . Fix one of those so they match!
I describe the issue, along with sample code in a post on the forums here from a couple of weeks ago:
https://discuss.huggingface.co/t/decicoder-finetune-error-understanding-naive-attention-prefill/51998
Thank you for pointing it out.
The config file has been adjusted to resolve the issue.
Please give it another try :)
Thank you, it works now!