FineTranslations gaz_Latn

A bounded Oromo (gaz_Latn) to English translation workload derived from HuggingFaceFW/finetranslations, packaged for batched offline LLM inference experiments. The slice keeps rows whose full chat-template input length is input_tokens <= 10000. No source text is truncated. This preserves a meaningful single-language FineTranslations subset while removing only the longest tail that would dominate