Llama-3 models hallucinate when `stop` is not found in output.
## The problem
Issue togethercomputer/together-python#137 (closed, 3 comments): Llama-3 chat models on Together ignored the stop list: with stop=['test'] the model kept emitting repeated output ('bar foo foo fooassistant ...') instead of stopping. Llama-2 models did not have this problem.
## The verified resolution
Contributor orangetin explained the tokenizer ignores the end-of-turn token; adding '<|eot_id|>' to the stop list fixes it. Reporter Infrared1029 confirmed ('yeah works when i add it'). Contributor said they would also patch it server-side.
## Rule of thumb
always include <|eot_id|> in stop sequences for Llama-3 models on Together.