def base_architecture(args): args.encoder_pretrained_embed = getattr(args, "encoder_pretrained_embed", None) args.encoder_embed_dim = getattr(args, "encoder_embed_dim", 512) args.encoder_ffn_embed_dim = getattr(args, "encoder_ffn_embed_dim", 2048) args.encoder_layers = getattr(args, "encoder_layers", 6) args.encoder_attention_heads = getattr(args, "encoder_attention_heads", 8) args.encoder_freeze_embed = getattr(args, "encoder_freeze_embed", False) args.encoder_learned_pos = getattr(args, "encoder_learned_pos", False) args.encoder_normalize_before = getattr(args, "encoder_normalize_before", False) args.decoder_pretrained_embed = getattr(args, "decoder_pretrained_embed", None) args.decoder_embed_dim = getattr(args, "decoder_embed_dim", args.encoder_embed_dim) args.decoder_layers = getattr(args, "decoder_layers", 2) args.decoder_attention_heads = getattr(args, "decoder_attention_heads", 8) args.decoder_reduced_attention_dim = getattr( args, "decoder_reduced_attention_dim", None) args.decoder_lstm_units = getattr(args, "decoder_lstm_units", 512) args.decoder_out_embed_dim = getattr(args, "decoder_out_embed_dim", 256) args.decoder_freeze_embed = getattr(args, "decoder_freeze_embed", False) args.attention_dropout = getattr(args, "attention_dropout", 0.0) args.relu_dropout = getattr(args, "relu_dropout", 0.0) args.dropout = getattr(args, "dropout", 0.1) vocab_reduction.set_arg_defaults(args)
def base_architecture(args): args.encoder_pretrained_embed = getattr(args, "encoder_pretrained_embed", None) args.encoder_embed_dim = getattr(args, "encoder_embed_dim", 256) args.encoder_ffn_embed_dim = getattr(args, "encoder_ffn_embed_dim", 512) args.encoder_layers = getattr(args, "encoder_layers", 3) args.encoder_attention_heads = getattr(args, "encoder_attention_heads", 4) args.encoder_freeze_embed = getattr(args, "encoder_freeze_embed", False) args.encoder_learned_pos = getattr(args, "encoder_normalize_before", False) args.encoder_normalize_before = getattr(args, "encoder_learned_pos", False) args.decoder_pretrained_embed = getattr(args, "decoder_pretrained_embed", None) args.decoder_embed_dim = getattr(args, "decoder_embed_dim", args.encoder_embed_dim) args.decoder_ffn_embed_dim = getattr(args, "decoder_ffn_embed_dim", args.encoder_ffn_embed_dim) args.decoder_layers = getattr(args, "decoder_layers", 3) args.decoder_attention_heads = getattr(args, "decoder_attention_heads", 4) args.decoder_freeze_embed = getattr(args, "decoder_freeze_embed", False) args.decoder_learned_pos = getattr(args, "decoder_normalize_before", False) args.decoder_normalize_before = getattr(args, "decoder_learned_pos", False) args.share_decoder_input_output_embed = getattr( args, "share_decoder_input_output_embed", False) args.share_all_embeddings = getattr(args, "share_all_embeddings", False) args.attention_dropout = getattr(args, "attention_dropout", 0.) args.relu_dropout = getattr(args, "relu_dropout", 0.) args.dropout = getattr(args, "dropout", 0.1) args.adaptive_softmax_cutoff = getattr(args, "adaptive_softmax_cutoff", None) vocab_reduction.set_arg_defaults(args)
def base_architecture(args): # default architecture args.encoder_embed_dim = getattr(args, "encoder_embed_dim", 512) args.encoder_layers = getattr(args, "encoder_layers", 1) args.encoder_hidden_dim = getattr(args, "encoder_hidden_dim", 512) args.encoder_bidirectional = getattr(args, "encoder_bidirectional", False) args.encoder_dropout_in = getattr(args, "encoder_dropout_in", args.dropout) args.encoder_dropout_out = getattr(args, "encoder_dropout_out", args.dropout) args.decoder_embed_dim = getattr(args, "decoder_embed_dim", 512) args.decoder_layers = getattr(args, "decoder_layers", 1) args.decoder_hidden_dim = getattr(args, "decoder_hidden_dim", 512) args.decoder_out_embed_dim = getattr(args, "decoder_out_embed_dim", 512) args.attention_type = getattr(args, "attention_type", "dot") args.decoder_dropout_in = getattr(args, "decoder_dropout_in", args.dropout) args.decoder_dropout_out = getattr(args, "decoder_dropout_out", args.dropout) args.averaging_encoder = getattr(args, "averaging_encoder", False) args.encoder_freeze_embed = getattr(args, "encoder_freeze_embed", False) args.decoder_freeze_embed = getattr(args, "decoder_freeze_embed", False) args.ngram_decoder = getattr(args, "ngram_decoder", None) args.cell_type = getattr(args, "cell_type", "lstm") args.ngram_activation_type = getattr(args, "ngram_activation_type", "relu") vocab_reduction.set_arg_defaults(args) word_dropout.set_arg_defaults(args) args.sequence_lstm = getattr(args, "sequence_lstm", False) args.add_encodercoder_output_as_decoder_input = getattr( args, "add_encoder_output_as_decoder_input", False)
def base_architecture(args): # default architecture args.encoder_embed_dim = getattr(args, "encoder_embed_dim", 512) args.encoder_layers = getattr(args, "encoder_layers", 1) args.encoder_hidden_dim = getattr(args, "encoder_hidden_dim", 512) args.encoder_bidirectional = getattr(args, "encoder_bidirectional", False) args.encoder_dropout_in = getattr(args, "encoder_dropout_in", args.dropout) args.encoder_dropout_out = getattr(args, "encoder_dropout_out", args.dropout) args.decoder_embed_dim = getattr(args, "decoder_embed_dim", 512) args.decoder_layers = getattr(args, "decoder_layers", 1) args.decoder_hidden_dim = getattr(args, "decoder_hidden_dim", 512) args.decoder_out_embed_dim = getattr(args, "decoder_out_embed_dim", 512) args.attention_type = getattr(args, "attention_type", "dot") args.decoder_dropout_in = getattr(args, "decoder_dropout_in", args.dropout) args.decoder_dropout_out = getattr(args, "decoder_dropout_out", args.dropout) args.averaging_encoder = getattr(args, "averaging_encoder", False) args.encoder_freeze_embed = getattr(args, "encoder_freeze_embed", False) args.decoder_freeze_embed = getattr(args, "decoder_freeze_embed", False) args.ngram_decoder = getattr(args, "ngram_decoder", None) args.multi_encoder = getattr(args, "multi_encoder", None) args.multi_decoder = getattr(args, "multi_decoder", None) args.multi_decoder_is_lm = getattr(args, "multi_decoder_is_lm", None) args.multiling_encoder_lang = getattr(args, "multiling_encoder_lang", None) args.multi_model_training_schedule = getattr( args, "multi_model_training_schedule", "complete" ) args.multi_model_fixed_weights = getattr(args, "multi_model_fixed_weights", None) args.cell_type = getattr(args, "cell_type", "lstm") args.ngram_activation_type = getattr(args, "ngram_activation_type", "relu") vocab_reduction.set_arg_defaults(args) word_dropout.set_arg_defaults(args) args.sequence_lstm = getattr(args, "sequence_lstm", False) args.decoder_tie_embeddings = getattr(args, "decoder_tie_embeddings", False)
def base_architecture(args): # default architecture args.encoder_embed_dim = getattr(args, 'encoder_embed_dim', 512) args.encoder_layers = getattr(args, 'encoder_layers', 1) args.encoder_hidden_dim = getattr(args, 'encoder_hidden_dim', 512) args.encoder_bidirectional = getattr(args, 'encoder_bidirectional', False) args.encoder_dropout_in = getattr(args, 'encoder_dropout_in', args.dropout) args.encoder_dropout_out = getattr(args, 'encoder_dropout_out', args.dropout) args.decoder_embed_dim = getattr(args, 'decoder_embed_dim', 512) args.decoder_layers = getattr(args, 'decoder_layers', 1) args.decoder_hidden_dim = getattr(args, 'decoder_hidden_dim', 512) args.decoder_out_embed_dim = getattr(args, 'decoder_out_embed_dim', 512) args.attention_type = getattr(args, 'attention_type', 'dot') args.decoder_dropout_in = getattr(args, 'decoder_dropout_in', args.dropout) args.decoder_dropout_out = getattr(args, 'decoder_dropout_out', args.dropout) args.averaging_encoder = getattr(args, 'averaging_encoder', False) args.encoder_freeze_embed = getattr(args, 'encoder_freeze_embed', False) args.decoder_freeze_embed = getattr(args, 'decoder_freeze_embed', False) args.cell_type = getattr(args, 'cell_type', 'lstm') vocab_reduction.set_arg_defaults(args) word_dropout.set_arg_defaults(args) args.sequence_lstm = getattr(args, 'sequence_lstm', False) args.add_encoder_output_as_decoder_input = getattr( args, 'add_encoder_output_as_decoder_input', False, )
def base_architecture(args): args.encoder_embed_path = getattr(args, "encoder_embed_path", None) args.encoder_embed_dim = getattr(args, "encoder_embed_dim", 512) args.encoder_ffn_embed_dim = getattr(args, "encoder_ffn_embed_dim", 2048) args.encoder_layers = getattr(args, "encoder_layers", 6) args.encoder_attention_heads = getattr(args, "encoder_attention_heads", 8) args.encoder_normalize_before = getattr(args, "encoder_normalize_before", False) args.encoder_learned_pos = getattr(args, "encoder_learned_pos", False) args.decoder_embed_path = getattr(args, "decoder_embed_path", None) args.decoder_embed_dim = getattr(args, "decoder_embed_dim", args.encoder_embed_dim) args.decoder_ffn_embed_dim = getattr(args, "decoder_ffn_embed_dim", args.encoder_ffn_embed_dim) args.decoder_layers = getattr(args, "decoder_layers", 6) args.decoder_attention_heads = getattr(args, "decoder_attention_heads", 8) args.decoder_normalize_before = getattr(args, "decoder_normalize_before", False) args.decoder_learned_pos = getattr(args, "decoder_learned_pos", False) args.attention_dropout = getattr(args, "attention_dropout", 0.0) args.relu_dropout = getattr(args, "relu_dropout", 0.0) args.dropout = getattr(args, "dropout", 0.1) args.adaptive_softmax_cutoff = getattr(args, "adaptive_softmax_cutoff", None) args.adaptive_softmax_dropout = getattr(args, "adaptive_softmax_dropout", 0) args.share_decoder_input_output_embed = getattr( args, "share_decoder_input_output_embed", False) args.share_all_embeddings = getattr(args, "share_all_embeddings", False) args.no_token_positional_embeddings = getattr( args, "no_token_positional_embeddings", False) args.decoder_output_dim = getattr(args, "decoder_output_dim", args.decoder_embed_dim) args.decoder_input_dim = getattr(args, "decoder_input_dim", args.decoder_embed_dim) args.decoder_aan_ffn = getattr(args, "decoder_aan_ffn", True) args.decoder_aan_ffn_use_embed_dim = getattr( args, "decoder_aan_ffn_use_embed_dim", False) args.decoder_aan_gating = getattr(args, "decoder_aan_gating", True) args.decoder_aan_more_dropouts = getattr(args, "decoder_aan_more_dropouts", "") vocab_reduction.set_arg_defaults(args)