cat((torch.arange(input_dim, dtype=torch.long).reshape(1, -1), self.h_[0].reshape(1, -1)), dim=0) indices2 = torch.cat((torch.arange(input_dim, dtype=torch.long).reshape(1, -1), self.h_[1].reshape(1, -1)), dim=0) self.sparseM = [ torch.sparse.FloatTensor(indices1,...
view(-1)# [BM,]pred_cls=torch.cat(outputs['pred_cls'],dim=1).view(-1,self.num_classes)# [BM, C]pred_box=torch.cat(outputs['pred_box'],dim=1).view(-1,4)# [BM, 4]# --- 标签分配 ---gt_objectness,gt_classes,gt_bboxes=self.matcher(fmp_sizes=fmp_sizes,fpn_strides=fpn_...
importtorch.nnasnnimporttorch.nn.functionalasFimporttorchclassVAE(nn.Module):def__init__(self):super().__init__()self.tok_emb=nn.Embedding(128,embedding_dim=128)self.cross_attn=MultiheadAttention(128,8)defforward(self,tokens):embed=self.tok_emb(tokens)z=torch.randn(embed.shape[0],4,emb...
cat([row, torch.full((num_edges, ), i, dtype=torch.long)]) choice = np.random.choice(torch.cat([row, col]).numpy(), num_edges) col = torch.cat([col, torch.from_numpy(choice)]) edge_index = torch.stack([row, col], dim=0) edge_index, _ = remove_self_loops(edge_index) ...
x_graph_vecs = self.W_graph( torch.cat([x_graph_vecs, diff_graph_vecs], dim=-1) ) loss, wacc, iacc, tacc, sacc = self.decoder((x_root_vecs, x_tree_vecs, x_graph_vecs), y_graphs, y_tensors, y_orders)returnloss + beta * kl_div, kl_div.item(), wacc, iacc, tacc, sac...
if num_valid < batch_size: zeros = stacked_sequence_output.data.new(num_layers, batch_size - num_valid, returned_timesteps, encoder_dim).fill_(0) zeros = Variable(zeros) stacked_sequence_output = torch.cat([stacked_sequence_output, zeros], 1) # The states also need to have invalid ...
size)x,t=measure_time(cat, [x1,x2],dim=1)# your function above# numpyx1,x2=np.random.randn(1,size),np.random.randn(1,size)x,t=measure_time(np.concatenate, [x1,x2],dim=1)# torchx1,x2=torch.randn(1,size),torch.randn(1,size)x,t=measure_time(torch.cat, [x1,x2],dim=1...
tokens = torch.cat((batch_pos, batch_neg),0) lengths = get_lengths(tokens, eos_idx) styles = torch.cat((pos_styles, neg_styles),0)returntokens, lengths, styles 开发者ID:plkmo,项目名称:NLP_Toolkit,代码行数:24,代码来源:train.py ...
= -1) if args.distributed: torch.cuda.set_device(args.local_rank) args.device = torch.device("cuda", args.local_rank) torch.distributed.init_process_group(backend='nccl', init_method='env://') logger.info( "Prepare tokenizer, pretraine...
cls_token = self.class_emb.expand(x.shape[0], -1, -1) x = torch.cat((cls_token, x), dim=1) if self.positional_emb is not None: x += self.positional_emb x = self.dropout(x) for blk in self.blocks: x = blk(x) x = self.norm(x) if self.seq_pool: x = torch.matmul...