Text Generation
Transformers
PyTorch
TensorBoard
Safetensors
bloom
Eval Results
text-generation-inference
Inference Endpoints

Output getting truncated while using Langchain

#247
by hepbc - opened

Hi folks: am using bloom with Langchain, initialized with temp=1e-10. I am using a FewShotPromptTemplate with 4 examples. It is doing an ok job but truncates the output at a random point. When I initialize with max_tokens, it gives the full string but fills up the balance of the output with arbitrary characters. Anyone else seen this happen and has a solution? Am using the free tier of HuggingFaceHub.

Thanks!

I do have the same issue, any luck with the parameters and testing?

Sign up or log in to comment