Open Access

Table 1

Best cross entropy test loss obtained for different architectures of the generative model.

Embedding Blocks Attention heads Positional embedding Best loss
64 4 4 No 2.422051429748535
64 4 4 Yes 2.445077657699585
128 6 8 No 2.4423086643218994
128 6 8 Yes 2.429002523422241
32 3 2 Yes 2.412593126296997
32 3 2 No 2.402848482131958
32 2 2 No 2.4102702140808105
64 3 2 No 2.4219555854797363
16 3 2 No 2.425621509552002
32 3 4 No 2.402952194213867
32 3 1 No 2.40149188041687

Notes. Our nominal and best model is indicated in bold face.

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.