試験NCA-GENM トピック1 問題33 スレッド

NVIDIA NCA-GENMのリアル試験問題集
問題 #: 33
トピック #: 1

You're using a pre-trained multimodal model that combines visual and textual information for a new downstream task: generating marketing slogans for product images. The model performs poorly, generating generic slogans that are unrelated to the specific product features. What is the MOST effective strategy to adapt this pre-trained model to your specific task?

A. Replace the model's output layer with a new layer trained specifically to generate marketing slogans.

B. Fine-tune the entire pre-trained model on a dataset of product images and corresponding marketing slogans.

C. Freeze the pre-trained model's weights and train a separate model to map the pre-trained model's output to marketing slogans.

D. Only fine-tune the visual encoder component of the pre-trained model.

E. Use the pre-trained model as is, without any adaptation.

おすすめの解答：B 解答を投票する

Fine-tuning the entire pre-trained model (B) allows the model to learn the specific nuances of the new task while leveraging the knowledge it gained during pre-training. Replacing only the output layer (A) might not be sufficient. Freezing the pre-trained model (C) limits its ability to adapt to the new task. Only fine-tuning the visual encoder (D) might not address the language generation aspect. Using the model without adaptation (E) will likely result in poor performance.

枪田** 2026-04-11 08:47:48

時間限定で

15%

オフ

プレミアムのNCA-GENM問題集のセルフテストエンジンまたはPDFをゲットしよう

弊社を連絡する

我々は１２時間以内ですべてのお問い合わせを答えます。

オンラインサポート時間：( UTC+9 ) 9:00-24:00
月曜日から土曜日まで

サポート：現在連絡

試験NCA-GENM トピック1 問題33 スレッド

コメント

弊社を連絡する

関連リンク

トップ試験