torch-rechub 0.0.1__py3-none-any.whl → 0.0.4__py3-none-any.whl

torch_rechub/trainers/trainer.py

@@ -1,111 +0,0 @@
-import os
-import torch
-import tqdm
-from sklearn.metrics import roc_auc_score
-from ..basic.callback import EarlyStopper
-
-
-class CTRTrainer(object):
-    """A general trainer for single task learning.
-
-    Args:
-        model (nn.Module): any multi task learning model.
-        optimizer_fn (torch.optim): optimizer function of pytorch (default = `torch.optim.Adam`).
-        optimizer_params (dict): parameters of optimizer_fn.
-        scheduler_fn (torch.optim.lr_scheduler) : torch scheduling class, eg. `torch.optim.lr_scheduler.StepLR`.
-        scheduler_params (dict): parameters of optimizer scheduler_fn.
-        n_epoch (int): epoch number of training.
-        earlystop_patience (int): how long to wait after last time validation auc improved (default=10).
-        device (str): `"cpu"` or `"cuda:0"`
-        gpus (list): id of multi gpu (default=[]). If the length >=1, then the model will wrapped by nn.DataParallel.
-        model_path (str): the path you want to save the model (default="./"). Note only save the best weight in the validation data.
-    """
-
-    def __init__(
-        self,
-        model,
-        optimizer_fn=torch.optim.Adam,
-        optimizer_params={
-            "lr": 1e-3,
-            "weight_decay": 1e-5
-        },
-        scheduler_fn=None,
-        scheduler_params=None,
-        n_epoch=10,
-        earlystop_patience=10,
-        device="cpu",
-        gpus=[],
-        model_path="./",
-    ):
-        self.model = model  #for uniform weights save method in one gpu or multi gpu
-        self.optimizer = optimizer_fn(self.model.parameters(), **optimizer_params)  #default optimizer
-        self.scheduler = None
-        if scheduler_fn is not None:
-            self.scheduler = scheduler_fn(self.optimizer, **scheduler_params)
-        self.criterion = torch.nn.BCELoss()  #default loss cross_entropy
-        self.evaluate_fn = roc_auc_score  #default evaluate function
-        self.n_epoch = n_epoch
-        self.early_stopper = EarlyStopper(patience=earlystop_patience)
-        self.device = torch.device(device)  #torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
-        self.gpus = gpus
-        if len(gpus) > 1:
-            print('parallel running on these gpus:', gpus)
-            self.model = torch.nn.DataParallel(self.model, device_ids=gpus)
-        self.model_path = model_path
-
-    def train_one_epoch(self, data_loader, log_interval=10):
-        self.model.train()
-        total_loss = 0
-        tk0 = tqdm.tqdm(data_loader, desc="train", smoothing=0, mininterval=1.0)
-        for i, (x_dict, y) in enumerate(tk0):
-            x_dict = {k: v.to(self.device) for k, v in x_dict.items()}  #tensor to GPU
-            y = y.to(self.device)
-            y_pred = self.model(x_dict)
-            loss = self.criterion(y_pred, y.float())
-            self.model.zero_grad()
-            loss.backward()
-            self.optimizer.step()
-            total_loss += loss.item()
-            if (i + 1) % log_interval == 0:
-                tk0.set_postfix(loss=total_loss / log_interval)
-                total_loss = 0
-
-    def fit(self, train_dataloader, val_dataloader):
-        self.model.to(self.device)
-        for epoch_i in range(self.n_epoch):
-            self.train_one_epoch(train_dataloader)
-            if self.scheduler is not None:
-                if epoch_i % self.scheduler.step_size == 0:
-                    print("Current lr : {}".format(self.optimizer.state_dict()['param_groups'][0]['lr']))
-                self.scheduler.step()  #update lr in epoch level by scheduler
-            auc = self.evaluate(self.model, val_dataloader)
-            print('epoch:', epoch_i, 'validation: auc:', auc)
-            if self.early_stopper.stop_training(auc, self.model.state_dict()):
-                print(f'validation: best auc: {self.early_stopper.best_auc}')
-                self.model.load_state_dict(self.early_stopper.best_weights)
-                torch.save(self.early_stopper.best_weights, os.path.join(self.model_path, "model.pth"))  #save best auc model
-                break
-
-    def evaluate(self, model, data_loader):
-        model.eval()
-        targets, predicts = list(), list()
-        with torch.no_grad():
-            tk0 = tqdm.tqdm(data_loader, desc="validation", smoothing=0, mininterval=1.0)
-            for i, (x_dict, y) in enumerate(tk0):
-                x_dict = {k: v.to(self.device) for k, v in x_dict.items()}
-                y = y.to(self.device)
-                y_pred = model(x_dict)
-                targets.extend(y.tolist())
-                predicts.extend(y_pred.tolist())
-        return self.evaluate_fn(targets, predicts)
-
-    def predict(self, model, data_loader):
-        model.eval()
-        predicts = list()
-        with torch.no_grad():
-            tk0 = tqdm.tqdm(data_loader, desc="predict", smoothing=0, mininterval=1.0)
-            for i, x_dict in enumerate(tk0):
-                x_dict = {k: v.to(self.device) for k, v in x_dict.items()}
-                y_pred = model(x_dict)
-                predicts.extend(y_pred.tolist())
-        return predicts

torch_rechub-0.0.1.dist-info/METADATA

@@ -1,105 +0,0 @@
-Metadata-Version: 2.1
-Name: torch-rechub
-Version: 0.0.1
-Summary: A Lighting Pytorch Framework for Recommendation System, Easy-to-use and Easy-to-extend.
-Home-page: https://github.com/morningsky/Torch-RecHub
-Author: Mincai lai
-Author-email: 757387961@qq.com
-License: UNKNOWN
-Keywords: ctr,click through rate,deep learning,pytorch,recsys,recommendation
-Platform: all
-Classifier: Intended Audience :: Developers
-Classifier: Intended Audience :: Education
-Classifier: Intended Audience :: Science/Research
-Classifier: Operating System :: OS Independent
-Classifier: Programming Language :: Python :: 3
-Classifier: Programming Language :: Python :: 3.7
-Classifier: Programming Language :: Python :: 3.8
-Classifier: Programming Language :: Python :: 3.9
-Classifier: Topic :: Scientific/Engineering
-Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
-Classifier: Topic :: Software Development :: Libraries
-Classifier: Topic :: Software Development :: Libraries :: Python Modules
-Classifier: License :: OSI Approved :: MIT License
-Description-Content-Type: text/markdown
-License-File: LICENSE
-Requires-Dist: numpy (>=1.21.5)
-Requires-Dist: torch (>=1.7.0)
-Requires-Dist: pandas (>=1.0.5)
-Requires-Dist: tqdm (>=4.64.0)
-Requires-Dist: scikit-learn (>=0.23.2)
-
-# Torch-RecHub
-
-A Lighting Pytorch Framework for Recommendation Models, Easy-to-use and Easy-to-extend.
-
-## 安装
-
-```python
-pip install torch-rechub
-```
-
-## 主要特性
-
-- scikit-learn风格易用的API（fit、predict），即插即用
-- 训练过程与模型定义解耦，易拓展，可针对不同类型的模型设置不同的训练机制
-- 使用Pytorch原生Dataset、DataLoader，易修改，自定义数据
-- 高度模块化，支持常见Layer（MLP、FM、FFM、target-attention、self-attention、transformer等），容易调用组装成新模型
-- 支持常见排序模型（WideDeep、DeepFM、DIN、DCN、xDeepFM等）
-
-- [ ] 支持常见召回模型（DSSM、YoutubeDNN、MIND、SARSRec等）
-- 丰富的多任务学习支持
-  - SharedBottom、ESMM、MMOE、PLE、AITM等模型
-  - GradNorm、UWL等动态loss加权机制
-
-- 聚焦更生态化的推荐场景
-  - [ ] 冷启动
-  - [ ] 延迟反馈
-  - [ ] 去偏
-- [ ] 支持丰富的训练机制（对比学习、蒸馏学习等）
-
-- [ ] 第三方高性能开源Trainer支持（Pytorch Lighting等）
-- [ ] 更多模型正在开发中
-
-## 快速使用
-
-```python
-from torch_rechub.models import WideDeep, DeepFM, DIN
-from torch_rechub.trainers import CTRTrainer
-from torch_rechub.basic.utils import DataGenerator
-
-dg = DataGenerator(x, y)
-train_dataloader, val_dataloader, test_dataloader = dg.generate_dataloader()
-
-model = DeepFM(deep_features=deep_features, fm_features=fm_features, mlp_params={"dims": [256, 128], "dropout": 0.2, "activation": "relu"})
-
-ctr_trainer = CTRTrainer(model)
-ctr_trainer.fit(train_dataloader, val_dataloader)
-auc = ctr_trainer.evaluate(ctr_trainer.model, test_dataloader)
-
-
-```
-
-
-
-
-
-> **Note:** 
->
-> 所有模型均在大多数论文提及的多个知名公开数据集中测试，达到或者接近论文性能。
->
-> 使用案例：[Examples](./examples)
->
-> 每个数据集将会提供
->
-> - 一个使用脚本，包含样本生成、模型训练与测试，并提供一套测评所用参数。
-> - 一个预处理脚本，将原始数据进行预处理，转化成csv。
-> - 数据格式参考文件（100条）。
-> - 全量数据，统一的csv文件，提供高速网盘下载链接和原始数据链接。
-
-
-
-[初步规划TODO清单](https://user-images.githubusercontent.com/11856746/167436396-f9c5de5b-d341-4697-8b91-884d4ae552be.png)
-
-
-

torch_rechub-0.0.1.dist-info/RECORD

@@ -1,26 +0,0 @@
-torch_rechub/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-torch_rechub/basic/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-torch_rechub/basic/activation.py,sha256=aIvomPA2X00dl3svv4lU2a4TvvltWxuqy22oML0e5YI,1519
-torch_rechub/basic/callback.py,sha256=Z62CIrPF2axyBSTRUnbc--9ADH8oF9YLTvLtxzvsMYE,953
-torch_rechub/basic/features.py,sha256=T4mA8I4JgCFZR0BZLrLRxKUdXa0Aaum9tJVlN3pdk_Y,1924
-torch_rechub/basic/layers.py,sha256=VCiaK0G4asvvzH4f7TobA7tyVhzgmc3FzvFK783z5yI,9354
-torch_rechub/basic/utils.py,sha256=v1NGrWVdj_CZBA7bERPEWqX5TcnQm0rVMDebDGXH17s,6581
-torch_rechub/models/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-torch_rechub/models/matching/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-torch_rechub/models/multi_task/__init__.py,sha256=tfy0NEOZlSkpSv6GSM3IFxPFGTrV3lpEXPAIGfUfvOM,106
-torch_rechub/models/multi_task/esmm.py,sha256=h7xSOxegyxfic5VFCEXEXtJLXanuRNBZ_T_rMMKZyyM,2168
-torch_rechub/models/multi_task/mmoe.py,sha256=2O4qUcmCMUKUW2gES6R84B2NTuKulgo6V01umhXPRQk,2818
-torch_rechub/models/multi_task/ple.py,sha256=m7k2D146Qg2ppKDtbXsPDFwXtSTjpsYO2hAFbFue7zQ,6467
-torch_rechub/models/multi_task/shared_bottom.py,sha256=3sBN5oaM912vJ1jzCpzaScnILIxobIA566cvQGS-TWQ,1855
-torch_rechub/models/ranking/__init__.py,sha256=k8fO0qs0WdbMcbOft2aV7NhXkmOAmBVTrQ6I1IMX7q4,78
-torch_rechub/models/ranking/deepfm.py,sha256=s3dAf5CYfNVHI62wXR2CZujmZ9V3yBdOBG6Y1ZgKOQc,1774
-torch_rechub/models/ranking/din.py,sha256=7UxS4ufcAi8ladJY68d3WEbn4986zHre8LaB4s9OVSs,4500
-torch_rechub/models/ranking/widedeep.py,sha256=ZpjhmIrrLygThUNP4ZWplgSboizrQjioEdy3SmGt3pM,1634
-torch_rechub/trainers/__init__.py,sha256=-HYw87_OnPJtwyDSUFioBNDJ7aiMM-bZNmiqV9vbfpw,67
-torch_rechub/trainers/mtl_trainer.py,sha256=yD98xeQuaI6cY6sN3ytDqrl4OQX113UA2kstDLC1rT4,6942
-torch_rechub/trainers/trainer.py,sha256=J7GhFed1QzBsUKntJD8qA2gMEtxAnx7haznpiousIk0,5052
-torch_rechub-0.0.1.dist-info/LICENSE,sha256=uz-kPWsf6PAphHAysV_p4A_MUcogaJbNrjPlINUC1x4,1067
-torch_rechub-0.0.1.dist-info/METADATA,sha256=XOlqfEwVKBuTpR2uXoavaT1hWC8AksgTPNlYy8-rcrk,3622
-torch_rechub-0.0.1.dist-info/WHEEL,sha256=G16H4A3IeoQmnOrYV4ueZGKSjhipXx8zc8nu9FGlvMA,92
-torch_rechub-0.0.1.dist-info/top_level.txt,sha256=_RPvAnlLHdll9u2d74aFr_oOAj_NxgfLpH02Uifz_YY,13
-torch_rechub-0.0.1.dist-info/RECORD,,

torch_rechub-0.0.1.dist-info/top_level.txt

@@ -1 +0,0 @@
-torch_rechub