The model predicts the next token for half sentences. (Giving probability like GPT-2.) To use the code, first download the ".saftensor" model. Enter the full path of the weight you downloaded instead of the "model_path" variable in the code (use // instead of / or you will get a UTF encoding error.). The model was trained on 1 "i514500HX" CPU for "1" epoch with 100 Turkish sentences. The model is purely for testing purposes. The model is free. ALL RISKS ARE YOURS!!
Inference Providers
NEW
This model isn't deployed by any Inference Provider.
🙋
Ask for provider support
HF Inference deployability: The model has no library tag.