I used the same implementation of Athena-Engine. But got AssertionError: Wrong masked weights I did some quite trick and found that the unit test has wrong mask and output. I’m be able to by pass the unit test :smiley: but someone should really fix it.

Course 5 Week 4 assignment scaled_dot_product_attention_test error

truclk August 29, 2021, 3:02am 4

Nevermind
I used the code from TensorFlow

it’s (mask * -1e9) but in assignment it’s (1-mask) * - 1e9

Scaled_dot_product_attention(q, k, v, mask) function:

Topic		Replies	Views
A problem with the Programming Assignment: Transformers Architecture with TensorFlow Sequence Models week-4	1	36	November 15, 2024
Stuck in C5_W4_A1_Transformer_Subclass_v1 errors Sequence Models	10	999	March 4, 2022
Trouble with Sequence Models DeepLearning.AI class program Sequence Models week-4	22	53	December 27, 2024
Course 5 Week 4 Exercise 3 :AssertionError: Wrong unmasked weights Sequence Models	1	413	October 20, 2023
Debugging C5 W4 Ex3 Sequence Models	1	579	October 23, 2021