# Stack Overflow [pytorch]: Llama model generation always skips 1 token during inference — retained record

> Retained stackexchange record from stackexchange.

Canonical: https://highsignal.app/data/records/4fd4e14ae9a817f3

[← sources](https://highsignal.app/data)


source: [stackexchange](https://highsignal.app/data/stackexchange) published 2026-08-02 12:43 ingested 2026-08-11 06:54

[original source ↗](https://stackoverflow.com/questions/79989890/llama-model-generation-always-skips-1-token-during-inference)

## retained text



```
Score: 1 | Answers: 0 | Views: 120
Tags: python, pytorch, huggingface-transformers, training-data
Llama model generation always skips 1 token during inference
```



document fetched 2026-08-25 08:00
