def __init__(self,
d_model=256,
d_ffn=1024,
dropout=0.1,
activation="relu",
n_levels=4,
n_heads=8,
n_points=4,
checkpoint_ffn=False):
super().__init__()
# cross attention
self.cross_attn = MSDeformAttn(d_model, n_levels, n_heads, n_points)
self.dropout1 = nn.Dropout(dropout)
self.norm1 = nn.LayerNorm(d_model)
# self attention
self.self_attn = nn.MultiheadAttention(d_model,
n_heads,
dropout=dropout)
self.dropout2 = nn.Dropout(dropout)
self.norm2 = nn.LayerNorm(d_model)
# ffn
self.linear1 = nn.Linear(d_model, d_ffn)
self.activation = _get_activation_fn(activation)
self.dropout3 = nn.Dropout(dropout)
self.linear2 = nn.Linear(d_ffn, d_model)
self.dropout4 = nn.Dropout(dropout)
self.norm3 = nn.LayerNorm(d_model)
# use torch.utils.checkpoint.checkpoint to save memory
self.checkpoint_ffn = checkpoint_ffn
def __init__(self,
d_model=256,
d_ffn=1024,
dropout=0.1,
activation="relu",
n_levels=4,
n_heads=8,
n_points=4,
sigmoid_attn=False):
super().__init__()
# self attention
self.self_attn = MSDeformAttn(d_model,
n_levels,
n_heads,
n_points,
sigmoid_attn=sigmoid_attn)
self.dropout1 = nn.Dropout(dropout)
self.norm1 = nn.LayerNorm(d_model)
# ffn
self.linear1 = nn.Linear(d_model, d_ffn)
self.activation = _get_activation_fn(activation)
self.dropout2 = nn.Dropout(dropout)
self.linear2 = nn.Linear(d_ffn, d_model)
self.dropout3 = nn.Dropout(dropout)
self.norm2 = nn.LayerNorm(d_model)
def __init__(self, d_model=256, d_ffn=1024,
dropout=0.1, activation="relu",
n_levels=4, n_heads=8, n_points=4, self_cross=True, sigmoid_attn=False, local_self_attn=False):
super().__init__()
self.self_cross = self_cross
self.local_self_attn = local_self_attn
self.num_head = n_heads
# cross attention
self.cross_attn = MSDeformAttn(d_model, n_levels, n_heads, n_points, sigmoid_attn=sigmoid_attn)
self.dropout1 = nn.Dropout(dropout)
self.norm1 = nn.LayerNorm(d_model)
# self attention
self.self_attn = nn.MultiheadAttention(d_model, n_heads, dropout=dropout)
self.dropout2 = nn.Dropout(dropout)
self.norm2 = nn.LayerNorm(d_model)
# ffn
self.linear1 = nn.Linear(d_model, d_ffn)
self.activation = _get_activation_fn(activation)
self.dropout3 = nn.Dropout(dropout)
self.linear2 = nn.Linear(d_ffn, d_model)
self.dropout4 = nn.Dropout(dropout)
self.norm3 = nn.LayerNorm(d_model)
def __init__(self,
d_model=256,
d_ffn=1024,
dropout=0.1,
activation="relu",
n_levels=4,
n_heads=8,
n_points=4,
self_cross=True,
sigmoid_attn=False,
extra_track_attn=False):
super().__init__()
self.self_cross = self_cross
self.num_head = n_heads
# cross attention
self.cross_attn = MSDeformAttn(d_model,
n_levels,
n_heads,
n_points,
sigmoid_attn=sigmoid_attn)
self.dropout1 = nn.Dropout(dropout)
self.norm1 = nn.LayerNorm(d_model)
self.self_attn = nn.MultiheadAttention(d_model,
n_heads,
dropout=dropout)
self.dropout2 = nn.Dropout(dropout)
self.norm2 = nn.LayerNorm(d_model)
# ffn
self.linear1 = nn.Linear(d_model, d_ffn)
self.activation = _get_activation_fn(activation)
self.dropout3 = nn.Dropout(dropout)
self.linear2 = nn.Linear(d_ffn, d_model)
self.dropout4 = nn.Dropout(dropout)
self.norm3 = nn.LayerNorm(d_model)
# update track query_embed
self.extra_track_attn = extra_track_attn
if self.extra_track_attn:
print('Training with Extra Self Attention in Every Decoder.',
flush=True)
self.update_attn = nn.MultiheadAttention(d_model,
n_heads,
dropout=dropout)
self.dropout5 = nn.Dropout(dropout)
self.norm4 = nn.LayerNorm(d_model)
if self_cross:
print('Training with Self-Cross Attention.')
else:
print('Training with Cross-Self Attention.')
def __init__(self,
d_model=256, d_ffn=1024,
dropout=0.1, activation="relu",
n_levels=4, n_heads=8, n_points=4, QSAttn=False):
super().__init__()
self.QSAttn = QSAttn
if self.QSAttn:
# siamese attention
self.siamese_attn = SingleHeadSiameseAttention(d_model)
# self attention
self.self_attn = MSDeformAttn(d_model, n_levels, n_heads, n_points)
self.dropout1 = nn.Dropout(dropout)
self.norm1 = nn.LayerNorm(d_model)
# ffn
self.linear1 = nn.Linear(d_model, d_ffn)
self.activation = _get_activation_fn(activation)
self.dropout3 = nn.Dropout(dropout)
self.linear2 = nn.Linear(d_ffn, d_model)
self.dropout4 = nn.Dropout(dropout)
self.norm3 = nn.LayerNorm(d_model)