I have:
for l_idx, layer in enumerate(reconstruct_spect_model.transformer.layers_encoding): for attn_idx, attention_map in enumerate(layer