Getting error for tokenizer add_prefix_space = True

#2
by Hitish9 - opened

i am getting to change this variable in tokenizer
AssertionError: You need to instantiate LongformerTokenizerFast with add_prefix_space=True to use it with pretokenized inputs.
when i am using 4k model for inference.

Ihor Stepanov helped me with answer

You can use:

from gliner import GLiNER
model = GLiNER.from_pretrained("numind/NuNER_Zero-4k")
model.data_processor.transformer_tokenizer.add_prefix_space=True

What about the quality? I tried and see that it goes in wrong direction!

I have not tried it much. you can also increase context window size of NunerZeroshot by increase max_len value in model config

Sign up or log in to comment