In the func "_enas_cell" of /NAO-WS/cnn/model_search.py, “num_possible_inputs = curr_cell + 1”. When curr_cell=0, prev_cell can be 0 or 1, num_possible_inputs=curr_cell+1=1, so the shape of "w" will be (1, avg_pool_c * out_filters), the code "w[prev_cell]" may cause a mistake of "out of index".
And I see in the func "_enas_conv", num_possible_inputs = curr_cell + 2. The other parts of these two functions are similar. So i wonder whether "curr_cell+1" should be "curr_cell+2" ? Thank you!
In the func "_enas_cell" of /NAO-WS/cnn/model_search.py, “num_possible_inputs = curr_cell + 1”. When curr_cell=0, prev_cell can be 0 or 1, num_possible_inputs=curr_cell+1=1, so the shape of "w" will be (1, avg_pool_c * out_filters), the code "w[prev_cell]" may cause a mistake of "out of index". And I see in the func "_enas_conv", num_possible_inputs = curr_cell + 2. The other parts of these two functions are similar. So i wonder whether "curr_cell+1" should be "curr_cell+2" ? Thank you!