# ByT5 XL model. include 't5x/examples/t5/byt5/base.gin' # imports vocab, optimizer and model. # ------------------- Network specification overrides -------------------------- network.Transformer.config = @network.T5Config() network.T5Config: emb_dim = 2560 num_heads = 32 num_encoder_layers = 36 num_decoder_layers = 12 head_dim = 64 mlp_dim = 6720