HugeCTR 开源项目教程

HugeCTR 开源项目教程

HugeCTRHugeCTR is a high efficiency GPU framework designed for Click-Through-Rate (CTR) estimating training项目地址:https://gitcode.com/gh_mirrors/hu/HugeCTR

项目介绍

HugeCTR 是一个专为点击率（CTR）估计训练设计的高效 GPU 框架。它旨在加速大规模深度学习模型的训练和推理，特别适用于推荐系统领域。HugeCTR 的设计目标包括：

• 快速：在包括 MLPerf 在内的推荐基准测试中表现出色。
• 易用：无论您是数据科学家还是机器学习从业者，我们都使其易于使用，提供丰富的文档、笔记本和示例。
• 领域特定：提供必要的功能，以便您可以高效地部署大型推荐模型。

项目快速启动

安装 HugeCTR

首先，确保您的系统满足 HugeCTR 的依赖要求，然后通过以下命令克隆项目并安装：

git clone https://github.com/NVIDIA-Merlin/HugeCTR.git
cd HugeCTR
mkdir build
cd build
cmake ..
make -j

训练模型

以下是一个简单的示例，展示如何使用 HugeCTR 训练一个模型：

import hugectr
from mpi4py import MPI

# 初始化 MPI
comm = MPI.COMM_WORLD

# 创建 HugeCTR 会话
solver = hugectr.CreateSolver(max_eval_batches = 100,
                              batchsize_eval = 1000,
                              batchsize = 1000,
                              vvgpu = [[0]],
                              use_mixed_precision = False,
                              i64_input_key = True,
                              use_algorithm_search = True,
                              use_cuda_graph = True,
                              repeat_dataset = True)

# 加载数据
reader = hugectr.DataReaderParams(data_reader_type = hugectr.DataReaderType_t.Parquet,
                                  source = ["train.parquet"],
                                  eval_source = "eval.parquet",
                                  check_type = hugectr.Check_t.Non)

# 创建模型
optimizer = hugectr.CreateOptimizer(optimizer_type = hugectr.Optimizer_t.Adam,
                                    update_type = hugectr.Update_t.Global,
                                    beta1 = 0.9,
                                    beta2 = 0.999,
                                    epsilon = 0.0000001)

model = hugectr.Model(solver, reader, optimizer)
model.add(hugectr.Input(label_dim = 1, label_name = "label",
                        dense_dim = 13, dense_name = "dense",
                        data_reader_sparse_param_array = 
                        [hugectr.DataReaderSparseParam("data1", 1, True, 26)]))
model.add(hugectr.SparseEmbedding(embedding_type = hugectr.Embedding_t.DistributedSlotSparseEmbeddingHash, 
                                  workspace_size_per_gpu_in_mb = 28,
                                  embedding_vec_size = 16,
                                  combiner = "sum",
                                  sparse_embedding_name = "sparse_embedding1",
                                  bottom_name = "data1",
                                  optimizer = optimizer))
model.add(hugectr.DenseLayer(layer_type = hugectr.Layer_t.InnerProduct,
                             bottom_names = ["dense"],
                             top_names = ["fc1"],
                             num_output = 1024))
model.add(hugectr.DenseLayer(layer_type = hugectr.Layer_t.ReLU,
                             bottom_names = ["fc1"],
                             top_names = ["relu1"]))
model.add(hugectr.DenseLayer(layer_type = hugectr.Layer_t.InnerProduct,
                             bottom_names = ["relu1"],
                             top_names = ["fc2"],
                             num_output = 1024))
model.add(hugectr.DenseLayer(layer_type = hugectr.Layer_t.ReLU,
                             bottom_names = ["fc2"],
                             top_names = ["relu2"]))
model.add(hugectr.DenseLayer(layer_type = hugectr.Layer_t.InnerProduct,
                             bottom_names = ["relu2"],
                             top_names = ["fc3"],
                             num_output = 1))
model.add(hugectr.DenseLayer(layer_type = hugectr.Layer_t

HugeCTRHugeCTR is a high efficiency GPU framework designed for Click-Through-Rate (CTR) estimating training项目地址:https://gitcode.com/gh_mirrors/hu/HugeCTR