Can not run prediction for new datasets about textee HOT 4 CLOSED

Hello,

I'am trying to run this repo for evaluating it on my datasets. Could you tell me where to get [saved_models_folder] in two prediction commands at the section "Making Predictions for New Texts". Is it the "TextEE/models" folder? Or somewhere else?

Thanks.

[saved_model_folder] is the saved checkpoint during training. Currently, we do not release any pre-trained models. Therefore, you may choose to train your own model with the command here. After training, [saved_model_folder] should be set as the folder under the output_dir in the config file. An example is here.

from textee.

Btw, I am working in the Kaggle notebook. The command I use the prediction command as below.

!/opt/conda/envs/textee/bin/python3.8 TextEE/predict_pipeline.py --input_file demo_input.txt --ed_model TextEE/models --eae_model TextEE/models --output_file demo_output.json

And I get this error:

Could you show me where it not working?

Thanks for pointing out. We forgot to remove it in our internal test for LLMs. We've fixed it now.

from textee.

Btw, I am working in the Kaggle notebook. The command I use the prediction command as below.

!/opt/conda/envs/textee/bin/python3.8 TextEE/predict_pipeline.py --input_file demo_input.txt --ed_model TextEE/models --eae_model TextEE/models --output_file demo_output.json

And I get this error:

Could you show me where it not working?

from textee.

Hello,
I'am trying to run this repo for evaluating it on my datasets. Could you tell me where to get [saved_models_folder] in two prediction commands at the section "Making Predictions for New Texts". Is it the "TextEE/models" folder? Or somewhere else?
Thanks.

[saved_model_folder] is the saved checkpoint during training. Currently, we do not release any pre-trained models. Therefore, you may choose to train your own model with the command here. After training, [saved_model_folder] should be set as the folder under the output_dir in the config file. An example is here.

I got it. Thank you for your reply.

from textee.