model_configs/custom_transformer.py

from fairseq.models import register_model_architecture
from fairseq.models.transformer import base_architecture


@register_model_architecture("transformer", "transformer_2x")
def transformer_big(args):
    args.encoder_embed_dim = getattr(args, "encoder_embed_dim", 1024)
    args.encoder_ffn_embed_dim = getattr(args, "encoder_ffn_embed_dim", 4096)
    args.encoder_attention_heads = getattr(args, "encoder_attention_heads", 16)
    args.encoder_normalize_before = getattr(args, "encoder_normalize_before", False)
    args.decoder_embed_dim = getattr(args, "decoder_embed_dim", 1024)
    args.decoder_ffn_embed_dim = getattr(args, "decoder_ffn_embed_dim", 4096)
    args.decoder_attention_heads = getattr(args, "decoder_attention_heads", 16)
    base_architecture(args)


@register_model_architecture("transformer", "transformer_4x")
def transformer_huge(args):
    args.encoder_embed_dim = getattr(args, "encoder_embed_dim", 1536)
    args.encoder_ffn_embed_dim = getattr(args, "encoder_ffn_embed_dim", 4096)
    args.encoder_attention_heads = getattr(args, "encoder_attention_heads", 16)
    args.encoder_normalize_before = getattr(args, "encoder_normalize_before", False)
    args.decoder_embed_dim = getattr(args, "decoder_embed_dim", 1536)
    args.decoder_ffn_embed_dim = getattr(args, "decoder_ffn_embed_dim", 4096)
    args.decoder_attention_heads = getattr(args, "decoder_attention_heads", 16)
    base_architecture(args)


@register_model_architecture("transformer", "transformer_9x")
def transformer_xlarge(args):
    args.encoder_embed_dim = getattr(args, "encoder_embed_dim", 2048)
    args.encoder_ffn_embed_dim = getattr(args, "encoder_ffn_embed_dim", 8192)
    args.encoder_attention_heads = getattr(args, "encoder_attention_heads", 16)
    args.encoder_normalize_before = getattr(args, "encoder_normalize_before", False)
    args.decoder_embed_dim = getattr(args, "decoder_embed_dim", 2048)
    args.decoder_ffn_embed_dim = getattr(args, "decoder_ffn_embed_dim", 8192)
    args.decoder_attention_heads = getattr(args, "decoder_attention_heads", 16)
    base_architecture(args)