diff --git a/docs/how_to/relay/01-build_network.md b/docs/how_to/relay/01-build_network.md index 6efa26b4..363a5e4d 100644 --- a/docs/how_to/relay/01-build_network.md +++ b/docs/how_to/relay/01-build_network.md @@ -12,6 +12,11 @@ title: 构建图卷积网络 本文介绍如何用 Relay 构建图卷积网络（GCN）。本教程演示在 Cora 数据集上运行 GCN。Cora 数据集是图神经网络（GNN）的 benchmark，同时是支持 GNN 训练和推理的框架。我们直接从 DGL 库加载数据集来与 DGL 进行同类比较。 +```bash +pip install torch==2.0.0 +pip install dgl==v1.0.0 +``` + 有关 DGL 安装，参阅 [DGL 文档](https://docs.dgl.ai/install/index.html)。 有关 PyTorch 安装，参阅 [PyTorch 指南](https://pytorch.org/get-started/locally/)。 @@ -64,22 +69,12 @@ Using backend: pytorch from dgl.data import load_data from collections import namedtuple -def load_dataset(dataset="cora"): - args = namedtuple("args", ["dataset"]) - data = load_data(args(dataset)) - - # 删除自循环，避免重复将节点的特征传递给自身 - g = data.graph - g.remove_edges_from(nx.selfloop_edges(g)) - g.add_edges_from(zip(g.nodes, g.nodes)) - - return g, data - -def evaluate(data, logits): - test_mask = data.test_mask # 未包含在训练阶段的测试集 +def evaluate(g, logits): + label = g.ndata["label"] + test_mask = g.ndata["test_mask"] pred = logits.argmax(axis=1) - acc = ((pred == data.labels) * test_mask).sum() / test_mask.sum() + acc = (torch.Tensor(pred[test_mask]) == label[test_mask]).float().mean() return acc ``` @@ -90,9 +85,6 @@ def evaluate(data, logits): """ Parameters ---------- -dataset: str - Name of dataset. You can choose from ['cora', 'citeseer', 'pubmed']. - num_layer: int number of hidden layers @@ -106,13 +98,14 @@ num_classes: int dimension of model output (Number of classes) """ -dataset = "cora" -g, data = load_dataset(dataset) +dataset = dgl.data.CoraGraphDataset() +dgl_g = dataset[0] num_layers = 1 num_hidden = 16 -infeat_dim = data.features.shape[1] -num_classes = data.num_labels +features = dgl_g.ndata["feat"] +infeat_dim = features.shape[1] +num_classes = dataset.num_classes ``` 输出结果： @@ -143,16 +136,14 @@ Done saving data into cached files. ``` python from tvm.contrib.download import download_testdata -from dgl import DGLGraph -features = torch.FloatTensor(data.features) -dgl_g = DGLGraph(g) +features = torch.FloatTensor(features) torch_model = GCN(dgl_g, infeat_dim, num_hidden, num_classes, num_layers, F.relu) # 下载预训练的权重 -model_url = "https://homes.cs.washington.edu/~cyulin/media/gnn_model/gcn_%s.torch" % (dataset) -model_path = download_testdata(model_url, "gcn_%s.pickle" % (dataset), module="gcn_model") +model_url = "https://homes.cs.washington.edu/~cyulin/media/gnn_model/gcn_cora.torch" +model_path = download_testdata(model_url, "gcn_cora.pickle", module="gcn_model") # 将 weights 加载到模型中 torch_model.load_state_dict(torch.load(model_path)) @@ -177,7 +168,7 @@ with torch.no_grad(): logits_torch = torch_model(features) print("Print the first five outputs from DGL-PyTorch execution\n", logits_torch[:5]) -acc = evaluate(data, logits_torch.numpy()) +acc = evaluate(dgl_g, logits_torch.numpy()) print("Test accuracy of DGL results: {:.2%}".format(acc)) ``` @@ -270,26 +261,25 @@ def GraphConv(layer_name, input_dim, output_dim, adj, input, norm=None, bias=Tru import numpy as np import networkx as nx -def prepare_params(g, data): +def prepare_params(g): params = {} - params["infeats"] = data.features.numpy().astype( - "float32" - ) # 目前仅支持 float32 格式 + params["infeats"] = g.ndata["feat"].numpy().astype("float32") # 生成邻接矩阵 - adjacency = nx.to_scipy_sparse_matrix(g) + nx_graph = dgl.to_networkx(g) + adjacency = nx.to_scipy_sparse_array(nx_graph) params["g_data"] = adjacency.data.astype("float32") params["indices"] = adjacency.indices.astype("int32") params["indptr"] = adjacency.indptr.astype("int32") # 标准化 w.r.t.节点的度 - degs = [g.in_degree[i] for i in range(g.number_of_nodes())] + degs = [g.in_degrees(i) for i in range(g.number_of_nodes())] params["norm"] = np.power(degs, -0.5).astype("float32") params["norm"] = params["norm"].reshape((params["norm"].shape[0], 1)) return params -params = prepare_params(g, data) +params = prepare_params(dgl_g) # 检查特征的 shape 和邻接矩阵的有效性 assert len(params["infeats"].shape) == 2 @@ -310,7 +300,7 @@ assert params["infeats"].shape[0] == params["indptr"].shape[0] - 1 ``` python # 在 Relay 中定义输入特征、范数、邻接矩阵 -infeats = relay.var("infeats", shape=data.features.shape) +infeats = relay.var("infeats", shape=features.shape) norm = relay.Constant(tvm.nd.array(params["norm"])) g_data = relay.Constant(tvm.nd.array(params["g_data"])) indices = relay.Constant(tvm.nd.array(params["indices"])) @@ -391,10 +381,7 @@ m.run() logits_tvm = m.get_output(0).numpy() print("Print the first five outputs from TVM execution\n", logits_tvm[:5]) -labels = data.labels -test_mask = data.test_mask - -acc = evaluate(data, logits_tvm) +acc = evaluate(dgl_g, logits_tvm) print("Test accuracy of TVM results: {:.2%}".format(acc)) import tvm.testing diff --git a/docs/how_to/relay/03-pipeline.md b/docs/how_to/relay/03-pipeline.md index 97015030..e7c39247 100644 --- a/docs/how_to/relay/03-pipeline.md +++ b/docs/how_to/relay/03-pipeline.md @@ -21,12 +21,7 @@ from tvm.relay.op.contrib.cutlass import partition_for_cutlass from tvm import relay from tvm.relay import testing import tvm.testing -from tvm.contrib.cutlass import ( - has_cutlass, - num_cutlass_partitions, - finalize_modules, - finalize_modules_vm, -) +from tvm.contrib.cutlass import finalize_modules img_size = 8 ``` @@ -44,7 +39,6 @@ def get_network(): "dweight", relay.TensorType((batch_size, 16 * img_size * img_size), "float16") ) weight = relay.var("weight") - second_weight = relay.var("second_weight") bn_gamma = relay.var("bn_gamma") bn_beta = relay.var("bn_beta") bn_mmean = relay.var("bn_mean") diff --git a/docs/how_to/relay/04-relay_visualizer.md b/docs/how_to/relay/04-relay_visualizer.md index 6c6875a0..c11f186f 100644 --- a/docs/how_to/relay/04-relay_visualizer.md +++ b/docs/how_to/relay/04-relay_visualizer.md @@ -16,6 +16,12 @@ Relay IR 模块可以包含很多操作。通常单个操作很容易理解， 这里用渲染器来渲染文本形式的计算图，它是一个轻量级、类似 AST 可视化工具（灵感来自 [clang ast-dump](https://clang.llvm.org/docs/IntroductionToTheClangAST.html)）。以下将介绍如何通过接口类来实现自定义的解析器和渲染器。 +安装依赖可运行： + +```bash +pip install graphviz +``` + 更多细节参考 `tvm.contrib.relay_viz`。 ``` python