Тест за Qwen на Alibaba

Отговорите и обясненията се появяват, докато отговаряте на теста.

1. Коя компания разработва семейството Qwen модели с изкуствен интелект?
   • Alibaba
   • Tencent
   • Baidu
   • ByteDance
2. What should developers check before using particular Qwen weights commercially?
   • Only the Qwen source-code repository license
   • The license or agreement supplied with those specific weights
   • Whether the model is popular
   • The parameter count, which determines rights automatically
3. What total and activated parameter counts did Alibaba report for Qwen1.5-MoE-A2.7B?
   • 2.7B total and 14.3B activated
   • 72B total and 72B activated
   • 14.3B total and 2.7B activated
   • 14.3B total with no parameters activated per token
4. Кой специализиран вариант на Qwen е предназначен за задачи по програмиране?
   • Qwen-VL
   • Qwen-Audio
   • Qwen-Math
   • Qwen-Кодер
5. Qwen-VL се отнася за модел, който може да обработва коя комбинация?
   • Визия (образи) и език
   • Данни за глас и местоположение
   • Видео игри и логически пъзели
   • Само много големи електронни таблици
6. According to the Qwen technical report, which language focus did the authors identify in the pretraining data?
   • Only French and German
   • китайски и английски
   • Само езици за програмиране
   • A single unnamed language
7. What does Qwen2-VL’s Naive Dynamic Resolution mechanism do?
   • Forces every input image into one fixed 224-pixel square
   • Removes all text from images
   • Processes images at varying resolutions into varying numbers of visual tokens
   • Converts image pixels directly into audio tokens
8. In a Mixture-of-Experts model, what directs a token to selected experts?
   • A person choosing every expert
   • The tokenizer’s vocabulary order
   • A fixed round-robin schedule
   • A learned router