def reverse( self, clone=True ):
if clone:
rval = self.clone()
else:
rval = self
rval.sequence = transform.reverse( self.sequence )
rval.quality = rval.quality[::-1]
return rval
def reverse(self, clone=True):
if clone:
rval = self.clone()
else:
rval = self
rval.sequence = transform.reverse(self.sequence)
rval.quality = rval.quality[::-1]
return rval
def reverse( self, clone = True ):
#need to override how color space is reversed
if clone:
rval = self.clone()
else:
rval = self
if rval.has_adapter_base():
adapter = rval.sequence[0]
#sequence = rval.sequence[1:]
rval.sequence = self.color_space_converter.to_color_space( transform.reverse( self.color_space_converter.to_base_space( rval.sequence ) ), adapter_base = adapter )
else:
rval.sequence = transform.reverse( rval.sequence )
if rval.is_ascii_encoded():
rval.quality = rval.quality[::-1]
else:
rval.quality = reversed( rval.get_decimal_quality_scores() )
rval.quality = "%s " % " ".join( map( str, rval.quality ) )
return rval
def forward(self, batch, h, c):
"""Compute tree-lstm prediction given a batch.
Parameters
----------
batch : dgl.data.SSTBatch
The data batch.
h : Tensor
Initial hidden state.
c : Tensor
Initial cell state.
Returns
-------
out
"""
g = batch.graph
h_bottom_up = self.propagate(g, self.cell_bottom_up, batch.X, h, c)
g_rev = dgl.batch([reverse(gu) for gu in dgl.unbatch(batch.graph)])
h_top_down = self.propagate(g_rev, self.cell_top_down,
th.cat([batch.X, h_bottom_up], dim=1), h,
c)
# indexes of root nodes
root_ids = batch.isroot.nonzero().flatten()
# h of root nodes
root_h_bottom_up = th.index_select(h_bottom_up, 0, root_ids)
# limit of ids of trees in graphs batch
lims_ids = root_ids.tolist() + [g.number_of_nodes()]
trees_h = [
h_top_down[s:e, :] for s, e in zip(lims_ids[:-1], lims_ids[1:])
]
trees_isleaf = [
batch.isleaf[s:e] for s, e in zip(lims_ids[:-1], lims_ids[1:])
]
leaves_h_top_down = th.cat([
th.mean(th.index_select(tree, 0,
leaves.nonzero().flatten()),
dim=0).view(1, -1)
for (tree, leaves) in zip(trees_h, trees_isleaf)
],
dim=0)
# average of h of non root node by tree
#inner_h_top_down = th.cat([th.mean(h_top_down[s+1:e-1,:],dim=0).view(1,-1) for s, e in zip(lims_ids[:-1],lims_ids[1:])])
out = th.cat([root_h_bottom_up, leaves_h_top_down], dim=1)
#out = root_h_bottom_up
return out
def reverse( self, clone = True ):
#need to override how decimal quality scores are reversed
if clone:
rval = self.clone()
else:
rval = self
rval.sequence = transform.reverse( self.sequence )
if rval.is_ascii_encoded():
rval.quality = rval.quality[::-1]
else:
rval.quality = reversed( rval.get_decimal_quality_scores() )
rval.quality = "%s " % " ".join( map( str, rval.quality ) )
return rval
def reverse(self, clone=True):
"""return a new object with a reverses self.sequence strand
call using my_object.reverse(clone=False) if you wish to reverse and
return the current object.
"""
if clone:
rval = self.clone()
else:
rval = self
rval.sequence = transform.reverse(self.sequence)
rval.quality = rval.quality[::-1]
return rval
def reverse(self, clone = True):
"""return a new object with a reverses self.sequence strand
call using my_object.reverse(clone=False) if you wish to reverse and
return the current object.
"""
if clone:
rval = self.clone()
else:
rval = self
rval.sequence = transform.reverse(self.sequence)
rval.quality = rval.quality[::-1]
return rval