# Beam search error
**URL:**
**Author:**  [@Felipehonorato](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2FFelipehonorato)%5C
**Post date:** [April 26, 2021, 1:30pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fbeam-search-error%2F5773%2F1 "2021-04-26T13:30:37Z")
Well, i’ve trained a new model on GPT2 and added some tokens to control my generation. It all worked fine while i was using greedy search algorithm, but when i tried to switch for beam search algorithm it started to mess up my tokens, mixing it with regular words. Does anyone have a clue about why is that happening? And how can i solve it?
---
**Author:**  [@brando](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fbrando)%5C
**Post date:** [May 12, 2024, 12:17am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fbeam-search-error%2F5773%2F2 "2024-05-12T00:17:27Z")
related: [How to generate multiple text completions per prompt (like vLLM) using HuggingFace Transformers Pipeline without triggering an error?](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fhow-to-generate-multiple-text-completions-per-prompt-like-vllm-using-huggingface-transformers-pipeline-without-triggering-an-error%2F86297)
related: [machine learning - How to generate multiple text completions per prompt (like vLLM) using HuggingFace Transformers Pipeline without triggering an error? - Stack Overflow](https://stackoverflow.com/questions/78466376/how-to-generate-multiple-text-completions-per-prompt-like-vllm-using-huggingfa)
---
**Author:**  [@brando](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fbrando)%5C
**Post date:** [May 12, 2024, 12:17am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fbeam-search-error%2F5773%2F3 "2024-05-12T00:17:34Z")
did you find an answer?