Updated eos_token to <|end|>

#3
by grafail - opened

Wrong token causes issues with https://github.com/huggingface/text-generation-inference, as it cannot be easily overridden.

Hugging Face H4 org

Hi @grafail you can set stop_sequences=["<|end|>"] in the text-generation-python client, or stop=["<|end|>"] if you're using the endpoint directly. I don't want to tamper with the EOS token since it can causes issues in other downstream applications

Ready to merge
This branch is ready to get merged automatically.

Sign up or log in to comment