def __init__(self,
args,
shared_compressed_layer,
no_encoder_attn=False,
add_bias_kv=False,
add_zero_attn=False):
super().__init__(PrimerConfig.from_namespace(args),
shared_compressed_layer, no_encoder_attn, add_bias_kv,
add_zero_attn)
def __init__(self, args, shared_conv_layer):
# wrap in a list so it's not automatically registered by PyTorch
self.shared_conv_layer = [shared_conv_layer]
super().__init__(PrimerConfig.from_namespace(args))
#self.fc1 = self.build_fc1(
# args.compressed_dim,
# args.encoder.ffn_embed_dim,
# self.quant_noise,
# self.quant_noise_block_size,
#)
self.register_buffer("version", torch.tensor(2))
def build_model(cls, args, task):
"""Build a new model instance."""
# make sure all arguments are present in older models
base_architecture(args)
if args.encoder_layers_to_keep:
args.encoder_layers = len(args.encoder_layers_to_keep.split(","))
if args.decoder_layers_to_keep:
args.decoder_layers = len(args.decoder_layers_to_keep.split(","))
if getattr(args, "max_positions", None) is None:
args.max_positions = DEFAULT_MAX_SOURCE_POSITIONS
if getattr(args, "max_source_positions", None) is None:
args.max_source_positions = DEFAULT_MAX_SOURCE_POSITIONS
if getattr(args, "max_target_positions", None) is None:
args.max_target_positions = DEFAULT_MAX_TARGET_POSITIONS
src_dict, tgt_dict = task.source_dictionary, task.target_dictionary
if args.share_all_embeddings:
if src_dict != tgt_dict:
raise ValueError(
"--share-all-embeddings requires a joined dictionary")
if args.encoder_embed_dim != args.decoder_embed_dim:
raise ValueError(
"--share-all-embeddings requires --encoder-embed-dim to match --decoder-embed-dim"
)
if args.decoder_embed_path and (args.decoder_embed_path !=
args.encoder_embed_path):
raise ValueError(
"--share-all-embeddings not compatible with --decoder-embed-path"
)
args.share_decoder_input_output_embed = True
if getattr(args, "offload_activations", False):
args.checkpoint_activations = True # offloading implies checkpointing
if not args.share_all_embeddings:
args.min_params_to_wrap = getattr(args, "min_params_to_wrap",
DEFAULT_MIN_PARAMS_TO_WRAP)
cfg = PrimerConfig.from_namespace(args)
return super().build_model(cfg, task)
def build_decoder(cls, args, tgt_dict, embed_tokens):
return super().build_decoder(PrimerConfig.from_namespace(args),
tgt_dict, embed_tokens)
def build_embedding(cls, args, dictionary, embed_dim, path=None):
return super().build_embedding(PrimerConfig.from_namespace(args),
dictionary, embed_dim, path)
def __init__(self, args, encoder, decoder):
cfg = PrimerConfig.from_namespace(args)
super().__init__(cfg, encoder, decoder)
self.args = args
def __init__(self, args, shared_compressed_layer):
super().__init__(PrimerConfig.from_namespace(args),
shared_compressed_layer)