transformer and decoder
ده ماه پیش منتشر شده
تعداد بازدید: 111
کد پروژه: 470088
شرح پروژه
The goal of this assignment is for you to design a transformer model (Vaswani et al.), train it on a
machine translation task, and analyze the decoder block. Please note that you are expected to
write a detailed report.
• You will use the Multi30k dataset, which contains about 30,000 sentence pairs in English
and German, annotated with image descriptions. The dataset is available in the torchtext
library.
• The model should consist of an encoder and a decoder, each composed of multiple layers
of self-attention and feed-forward networks. You should also implement the positional
encoding, the scaled dot-product attention, and the multi-head attention mechanisms.
• You should monitor the training loss and the validation loss, and save the best model
based on the validation loss.
• You should also plot the attention maps of the decoder block, and interpret the results.
• evaluate the performance of the decoder block on a test set of sentence pairs, and to
report the metrics, such as the accuracy, the loss, and the BLEU score.
مهارت ها و تخصص های مورد نیاز
بودجه
750,000 تومان تا 5,000,000 تومان
مهلت برای انجام
9روز
وضعیت مناقصه
بسته
درباره کارفرما
عضویت ده ماه پیش
