Commit 7a36da42 authored by Angela Fan's avatar Angela Fan Committed by Myle Ott
Browse files

modified writing prompts model parameters to make readme cleaner

parent b59815bc

fairseq/models/fconv_self_att.py

+11 −3
Changes for fairseq/models/fconv_self_att.py: 11 added lines, 3 removed lines.
Original line number Diff line number Diff line
@@ -494,9 +494,17 @@ def base_architecture(args): 
@register_model_architecture('fconv_self_att', 'fconv_self_att_wp')

def fconv_self_att_wp(args):

    base_architecture(args)

    args.encoder_embed_dim = getattr(args, 'encoder_embed_dim', 512)

    args.encoder_embed_dim = getattr(args, 'encoder_embed_dim', 256)

    args.encoder_layers = getattr(args, 'encoder_layers', '[(128, 3)] * 2 + [(512,3)] * 1')

    args.decoder_embed_dim = getattr(args, 'decoder_embed_dim', 512)

    args.decoder_layers = getattr(args, 'decoder_layers', '[(512, 4)] * 4 + [(768, 4)] * 2 + [(1024, 4)] * 1 + [(2048,4)] * 1')

    args.decoder_embed_dim = getattr(args, 'decoder_embed_dim', 256)

    args.decoder_layers = getattr(args, 'decoder_layers', '[(512, 4)] * 4 + [(768, 4)] * 2 + [(1024, 4)] * 1')

    args.decoder_out_embed_dim = getattr(args, 'decoder_out_embed_dim', 256)

    args.decoder_attention = getattr(args, 'decoder_attention', 'True')

    args.multihead_attention_nheads = getattr(args, 'multihead_attention_nheads', 1)

    args.encoder_attention = getattr(args, 'encoder_attention', 'False')

    args.encoder_attention_nheads = getattr(args, 'encoder_attention_nheads', 1)

    args.self_attention = getattr(args, 'self_attention', 'True')

    args.gated_attention = getattr(args, 'gated_attention', 'True')

    args.downsample = getattr(args, 'downsample', 'True')

    args.project_input = getattr(args, 'project_input', 'True')

    args.multihead_self_attention_nheads = getattr(args, 'multihead_self_attention_nheads', 4)