diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml index fc3cdae4..4dad4ba1 100644 --- a/.github/workflows/ci.yml +++ b/.github/workflows/ci.yml @@ -17,19 +17,25 @@ jobs: timeout-minutes: 120 strategy: matrix: - python-version: [3.6,3.7] - torch-version: [1.1.0,1.2.0,1.3.0,1.4.0,1.5.0,1.6.0,1.7.0,1.8.1] + python-version: [3.6,3.7,3.8] + torch-version: [1.1.0,1.2.0,1.3.0,1.4.0,1.5.0,1.6.0,1.7.1,1.8.1,1.9.0,1.10.2,1.11.0] -# exclude: -# - python-version: 3.5 -# tf-version: 1.1.0 + exclude: + - python-version: 3.6 + torch-version: 1.11.0 + - python-version: 3.8 + torch-version: 1.1.0 + - python-version: 3.8 + torch-version: 1.2.0 + - python-version: 3.8 + torch-version: 1.3.0 steps: - - uses: actions/checkout@v1 + - uses: actions/checkout@v3 - name: Setup python environment - uses: actions/setup-python@v1 + uses: actions/setup-python@v4 with: python-version: ${{ matrix.python-version }} @@ -47,7 +53,7 @@ jobs: pip install -q sklearn pytest --cov=deepctr_torch --cov-report=xml - name: Upload coverage to Codecov - uses: codecov/codecov-action@v1.0.2 + uses: codecov/codecov-action@v3.1.0 with: token: ${{secrets.CODECOV_TOKEN}} file: ./coverage.xml diff --git a/README.md b/README.md index 7c646933..6d02554e 100644 --- a/README.md +++ b/README.md @@ -47,34 +47,19 @@ Let's [**Get Started!**](https://deepctr-torch.readthedocs.io/en/latest/Quick-St ## DisscussionGroup & Related Projects - - - - - - - -
- 公众号：浅梦学习笔记

- - - - 		- 微信：deepctrbot

- - - - 		- -
- +- [Github Discussions](https://github.com/shenweichen/DeepCTR/discussions) +- Wechat Discussions +|公众号：浅梦学习笔记|微信：deepctrbot|学习小组 [加入](https://t.zsxq.com/026UJEuzv) [主题集合](https://mp.weixin.qq.com/mp/appmsgalbum?__biz=MjM5MzY4NzE3MA==&action=getalbum&album_id=1361647041096843265&scene=126#wechat_redirect)| +|:--:|:--:|:--:| +| [![公众号](./docs/pics/code.png)](https://github.com/shenweichen/AlgoNotes)| [![微信](./docs/pics/deepctrbot.png)](https://github.com/shenweichen/AlgoNotes)|[![学习小组](./docs/pics/planet_github.png)](https://t.zsxq.com/026UJEuzv)| +- Related Projects + + - [AlgoNotes](https://github.com/shenweichen/AlgoNotes) + - [DeepCTR](https://github.com/shenweichen/DeepCTR) + - [DeepMatch](https://github.com/shenweichen/DeepMatch) + - [GraphEmbedding](https://github.com/shenweichen/GraphEmbedding) ## Main Contributors([welcome to join us!](./CONTRIBUTING.md)) @@ -84,59 +69,58 @@ Let's [**Get Started!**](https://deepctr-torch.readthedocs.io/en/latest/Quick-St ​ pic
​ Shen Weichen ​ -

Core Dev
Zhejiang Unversity

​ +

Alibaba Group

​ ​ pic
​ Zan Shuxun -

Core Dev
Beijing University
of Posts and
Telecommunications

​ +

Alibaba Group

​ pic
Wang Ze ​ -

Core Dev
Beihang University

​ +

Meituan

​ ​ pic
Zhang Wutong -

Core Dev
Beijing University
of Posts and
Telecommunications

​ +

Tencent

​ ​ pic
​ Zhang Yuefeng -

Core Dev
- Peking University

​ +

Peking University

​ ​ pic
​ Huo Junyi -

Core Dev
+

University of Southampton

​ ​ pic
​ Zeng Kai ​ -

Dev
+

SenseTime

​ ​ pic
​ Chen K ​ -

Dev
+

NetEase

​ ​ pic
​ Cheng Weiyu ​ -

Dev
+

Shanghai Jiao Tong University

​ ​ pic
​ Tang -

Test
+

Tongji University

​ diff --git a/deepctr_torch/__init__.py b/deepctr_torch/__init__.py index 88508515..4be7a5bc 100644 --- a/deepctr_torch/__init__.py +++ b/deepctr_torch/__init__.py @@ -2,5 +2,5 @@ from . import models from .utils import check_version -__version__ = '0.2.7' +__version__ = '0.2.8' check_version(__version__) \ No newline at end of file diff --git a/deepctr_torch/models/basemodel.py b/deepctr_torch/models/basemodel.py index 4235ad38..17e57b90 100644 --- a/deepctr_torch/models/basemodel.py +++ b/deepctr_torch/models/basemodel.py @@ -3,6 +3,7 @@ Author: Weichen Shen,weichenswc@163.com + zanshuxun, zanshuxun@aliyun.com """ from __future__ import print_function @@ -75,7 +76,7 @@ def forward(self, X, sparse_feat_refine_weight=None): sparse_embedding_list += varlen_embedding_list - linear_logit = torch.zeros([X.shape[0], 1]).to(sparse_embedding_list[0].device) + linear_logit = torch.zeros([X.shape[0], 1]).to(self.device) if len(sparse_embedding_list) > 0: sparse_embedding_cat = torch.cat(sparse_embedding_list, dim=-1) if sparse_feat_refine_weight is not None: @@ -476,6 +477,10 @@ def _log_loss(self, y_true, y_pred, eps=1e-7, normalize=True, sample_weight=None sample_weight, labels) + @staticmethod + def _accuracy_score(y_true, y_pred): + return accuracy_score(y_true, np.where(y_pred > 0.5, 1, 0)) + def _get_metrics(self, metrics, set_eps=False): metrics_ = {} if metrics: @@ -490,8 +495,7 @@ def _get_metrics(self, metrics, set_eps=False): if metric == "mse": metrics_[metric] = mean_squared_error if metric == "accuracy" or metric == "acc": - metrics_[metric] = lambda y_true, y_pred: accuracy_score( - y_true, np.where(y_pred > 0.5, 1, 0)) + metrics_[metric] = self._accuracy_score self.metrics_names.append(metric) return metrics_ diff --git a/docs/pics/code2.jpg b/docs/pics/code2.jpg new file mode 100644 index 00000000..e191f297 Binary files /dev/null and b/docs/pics/code2.jpg differ diff --git a/docs/pics/planet_github.png b/docs/pics/planet_github.png new file mode 100644 index 00000000..67efe968 Binary files /dev/null and b/docs/pics/planet_github.png differ diff --git a/docs/requirements.readthedocs.txt b/docs/requirements.readthedocs.txt index 80bc5e6d..793412bd 100644 --- a/docs/requirements.readthedocs.txt +++ b/docs/requirements.readthedocs.txt @@ -1,2 +1,3 @@ Cython>=0.28.5 -tensorflow==1.15.4 +tensorflow==2.7.2 +scikit-learn==1.0 diff --git a/docs/source/FAQ.md b/docs/source/FAQ.md index 102e35bc..1006bf42 100644 --- a/docs/source/FAQ.md +++ b/docs/source/FAQ.md @@ -6,7 +6,7 @@ To save/load weights: ```python import torch -model = DeepFM() +model = DeepFM(...) torch.save(model.state_dict(), 'DeepFM_weights.h5') model.load_state_dict(torch.load('DeepFM_weights.h5')) ``` @@ -15,7 +15,7 @@ To save/load models: ```python import torch -model = DeepFM() +model = DeepFM(...) torch.save(model, 'DeepFM.h5') model = torch.load('DeepFM.h5') ``` diff --git a/docs/source/History.md b/docs/source/History.md index ec68a102..4984dfc4 100644 --- a/docs/source/History.md +++ b/docs/source/History.md @@ -1,5 +1,6 @@ # History -- 06/14/2021 : [v0.2.7](https://github.com/shenweichen/DeepCTR-Torch/releases/tag/v0.2.6) released.Add [AFN](./Features.html#afn-adaptive-factorization-network-learning-adaptive-order-feature-interactions) and fix some bugs. +- 06/19/2022 : [v0.2.8](https://github.com/shenweichen/DeepCTR-Torch/releases/tag/v0.2.8) released.Fix some bugs. +- 06/14/2021 : [v0.2.7](https://github.com/shenweichen/DeepCTR-Torch/releases/tag/v0.2.7) released.Add [AFN](./Features.html#afn-adaptive-factorization-network-learning-adaptive-order-feature-interactions) and fix some bugs. - 04/04/2021 : [v0.2.6](https://github.com/shenweichen/DeepCTR-Torch/releases/tag/v0.2.6) released.Add [IFM](./Features.html#ifm-input-aware-factorization-machine) and [DIFM](./Features.html#difm-dual-input-aware-factorization-machine);Support multi-gpus running([example](./FAQ.html#how-to-run-the-demo-with-multiple-gpus)). - 02/12/2021 : [v0.2.5](https://github.com/shenweichen/DeepCTR-Torch/releases/tag/v0.2.5) released.Fix bug in DCN-M. - 12/05/2020 : [v0.2.4](https://github.com/shenweichen/DeepCTR-Torch/releases/tag/v0.2.4) released.Imporve compatibility & fix issues.Add History callback.([example](https://deepctr-torch.readthedocs.io/en/latest/FAQ.html#set-learning-rate-and-use-earlystopping)). diff --git a/docs/source/conf.py b/docs/source/conf.py index e99b48ea..615f48b0 100644 --- a/docs/source/conf.py +++ b/docs/source/conf.py @@ -26,7 +26,7 @@ # The short X.Y version version = '' # The full version, including alpha/beta/rc tags -release = '0.2.7' +release = '0.2.8' # -- General configuration --------------------------------------------------- diff --git a/docs/source/index.rst b/docs/source/index.rst index 1701d403..497d232b 100644 --- a/docs/source/index.rst +++ b/docs/source/index.rst @@ -34,18 +34,21 @@ You can read the latest code at https://github.com/shenweichen/DeepCTR-Torch and News ----- +06/19/2022 : Fix some bugs. `Changelog `_ + 06/14/2021 : Add `AFN <./Features.html#afn-adaptive-factorization-network-learning-adaptive-order-feature-interactions>`_ and fix some bugs. `Changelog `_ 04/04/2021 : Add `IFM <./Features.html#ifm-input-aware-factorization-machine>`_ and `DIFM <./Features.html#difm-dual-input-aware-factorization-machine>`_ . Support multi-gpus running(`example <./FAQ.html#how-to-run-the-demo-with-multiple-gpus>`_). `Changelog `_ -02/12/2021 : Fix bug in DCN-M. `Changelog `_ DisscussionGroup ----------------------- -公众号：**浅梦学习笔记** wechat ID: **deepctrbot** + 公众号：**浅梦学习笔记** wechat ID: **deepctrbot** + + `Discussions `_ `学习小组主题集合 `_ -.. image:: ../pics/code.png +.. image:: ../pics/code2.jpg .. toctree:: :maxdepth: 2 diff --git a/setup.py b/setup.py index 4e44fe13..705a9905 100644 --- a/setup.py +++ b/setup.py @@ -4,12 +4,12 @@ long_description = fh.read() REQUIRED_PACKAGES = [ - 'torch>=1.1.0', 'tqdm', 'sklearn', 'tensorflow' + 'torch>=1.1.0', 'tqdm', 'scikit-learn', 'tensorflow' ] setuptools.setup( name="deepctr-torch", - version="0.2.7", + version="0.2.8", author="Weichen Shen", author_email="weichenswc@163.com", description="Easy-to-use,Modular and Extendible package of deep learning based CTR(Click Through Rate) prediction models with PyTorch", @@ -37,6 +37,7 @@ 'Programming Language :: Python :: 3.5', 'Programming Language :: Python :: 3.6', 'Programming Language :: Python :: 3.7', + 'Programming Language :: Python :: 3.8', 'Topic :: Scientific/Engineering', 'Topic :: Scientific/Engineering :: Artificial Intelligence', 'Topic :: Software Development', diff --git a/tests/models/DeepFM_test.py b/tests/models/DeepFM_test.py index b11b5cb4..a11dc3bd 100644 --- a/tests/models/DeepFM_test.py +++ b/tests/models/DeepFM_test.py @@ -6,21 +6,33 @@ @pytest.mark.parametrize( - 'use_fm,hidden_size,sparse_feature_num', - [(True, (32,), 3), - (False, (32,), 3), - (False, (32,), 2), (False, (32,), 1), (True, (), 1), (False, (), 2) + 'use_fm,hidden_size,sparse_feature_num,dense_feature_num', + [(True, (32,), 3, 3), + (False, (32,), 3, 3), + (False, (32,), 2, 2), + (False, (32,), 1, 1), + (True, (), 1, 1), + (False, (), 2, 2), + (True, (32,), 0, 3), + (True, (32,), 3, 0), + (False, (32,), 0, 3), + (False, (32,), 3, 0), ] ) -def test_DeepFM(use_fm, hidden_size, sparse_feature_num): +def test_DeepFM(use_fm, hidden_size, sparse_feature_num, dense_feature_num): model_name = "DeepFM" sample_size = SAMPLE_SIZE x, y, feature_columns = get_test_data( - sample_size, sparse_feature_num=sparse_feature_num, dense_feature_num=sparse_feature_num) + sample_size, sparse_feature_num=sparse_feature_num, dense_feature_num=dense_feature_num) model = DeepFM(feature_columns, feature_columns, use_fm=use_fm, dnn_hidden_units=hidden_size, dnn_dropout=0.5, device=get_device()) check_model(model, model_name, x, y) + # no linear part + model = DeepFM([], feature_columns, use_fm=use_fm, + dnn_hidden_units=hidden_size, dnn_dropout=0.5, device=get_device()) + check_model(model, model_name + '_no_linear', x, y) + if __name__ == "__main__": pass diff --git a/tests/utils.py b/tests/utils.py index 10abcecb..28f3010b 100644 --- a/tests/utils.py +++ b/tests/utils.py @@ -4,6 +4,7 @@ import numpy as np import torch as torch +from deepctr_torch.callbacks import EarlyStopping, ModelCheckpoint from deepctr_torch.inputs import SparseFeat, DenseFeat, VarLenSparseFeat SAMPLE_SIZE = 64 @@ -17,46 +18,46 @@ def gen_sequence(dim, max_len, sample_size): def get_test_data(sample_size=1000, embedding_size=4, sparse_feature_num=1, dense_feature_num=1, sequence_feature=['sum', 'mean', 'max'], classification=True, include_length=False, hash_flag=False, prefix=''): - - feature_columns = [] model_input = {} - - if 'weight' in sequence_feature: - feature_columns.append(VarLenSparseFeat(SparseFeat(prefix+"weighted_seq",vocabulary_size=2,embedding_dim=embedding_size),maxlen=3,length_name=prefix+"weighted_seq"+"_seq_length",weight_name=prefix+"weight")) + if 'weight' in sequence_feature: + feature_columns.append( + VarLenSparseFeat(SparseFeat(prefix + "weighted_seq", vocabulary_size=2, embedding_dim=embedding_size), + maxlen=3, length_name=prefix + "weighted_seq" + "_seq_length", + weight_name=prefix + "weight")) s_input, s_len_input = gen_sequence( 2, 3, sample_size) - model_input[prefix+"weighted_seq"] = s_input - model_input[prefix+'weight'] = np.random.randn(sample_size,3,1) - model_input[prefix+"weighted_seq"+"_seq_length"] = s_len_input + model_input[prefix + "weighted_seq"] = s_input + model_input[prefix + 'weight'] = np.random.randn(sample_size, 3, 1) + model_input[prefix + "weighted_seq" + "_seq_length"] = s_len_input sequence_feature.pop(sequence_feature.index('weight')) - for i in range(sparse_feature_num): dim = np.random.randint(1, 10) - feature_columns.append(SparseFeat(prefix+'sparse_feature_'+str(i), dim,embedding_size,dtype=torch.int32)) + feature_columns.append(SparseFeat(prefix + 'sparse_feature_' + str(i), dim, embedding_size, dtype=torch.int32)) for i in range(dense_feature_num): - feature_columns.append(DenseFeat(prefix+'dense_feature_'+str(i), 1,dtype=torch.float32)) + feature_columns.append(DenseFeat(prefix + 'dense_feature_' + str(i), 1, dtype=torch.float32)) for i, mode in enumerate(sequence_feature): dim = np.random.randint(1, 10) maxlen = np.random.randint(1, 10) feature_columns.append( - VarLenSparseFeat(SparseFeat(prefix +'sequence_' + mode,vocabulary_size=dim, embedding_dim=embedding_size), maxlen=maxlen, combiner=mode)) + VarLenSparseFeat(SparseFeat(prefix + 'sequence_' + mode, vocabulary_size=dim, embedding_dim=embedding_size), + maxlen=maxlen, combiner=mode)) for fc in feature_columns: - if isinstance(fc,SparseFeat): - model_input[fc.name]= np.random.randint(0, fc.vocabulary_size, sample_size) - elif isinstance(fc,DenseFeat): + if isinstance(fc, SparseFeat): + model_input[fc.name] = np.random.randint(0, fc.vocabulary_size, sample_size) + elif isinstance(fc, DenseFeat): model_input[fc.name] = np.random.random(sample_size) else: s_input, s_len_input = gen_sequence( fc.vocabulary_size, fc.maxlen, sample_size) model_input[fc.name] = s_input if include_length: - fc.length_name = prefix+"sequence_"+str(i)+'_seq_length' - model_input[prefix+"sequence_"+str(i)+'_seq_length'] = s_len_input + fc.length_name = prefix + "sequence_" + str(i) + '_seq_length' + model_input[prefix + "sequence_" + str(i) + '_seq_length'] = s_len_input if classification: y = np.random.randint(0, 2, sample_size) @@ -66,7 +67,7 @@ def get_test_data(sample_size=1000, embedding_size=4, sparse_feature_num=1, dens return model_input, y, feature_columns -def layer_test(layer_cls, kwargs = {}, input_shape=None, +def layer_test(layer_cls, kwargs={}, input_shape=None, input_dtype=torch.float32, input_data=None, expected_output=None, expected_output_shape=None, expected_output_dtype=None, fixed_batch_size=False): '''check layer is valid or not @@ -90,7 +91,7 @@ def layer_test(layer_cls, kwargs = {}, input_shape=None, for i, e in enumerate(input_data_shape): if e is None: input_data_shape[i] = np.random.randint(1, 4) - + if all(isinstance(e, tuple) for e in input_data_shape): input_data = [] for e in input_data_shape: @@ -104,37 +105,37 @@ def layer_test(layer_cls, kwargs = {}, input_shape=None, # use input_data to update other parameters if input_shape is None: input_shape = input_data.shape - + if expected_output_dtype is None: expected_output_dtype = input_dtype - + # layer initialization layer = layer_cls(**kwargs) - + if fixed_batch_size: inputs = torch.tensor(input_data.unsqueeze(0), dtype=input_dtype) else: inputs = torch.tensor(input_data, dtype=input_dtype) - + # calculate layer's output output = layer(inputs) if not output.dtype == expected_output_dtype: raise AssertionError("layer output dtype does not match with the expected one") - + if not expected_output_shape: - raise ValueError("expected output shape should not be none") + raise ValueError("expected output shape should not be none") actual_output_shape = output.shape for expected_dim, actual_dim in zip(expected_output_shape, actual_output_shape): if expected_dim is not None: if not expected_dim == actual_dim: raise AssertionError(f"expected_dim:{expected_dim}, actual_dim:{actual_dim}") - + if expected_output is not None: # check whether output equals to expected output assert_allclose(output, expected_output, rtol=1e-3) - + return output @@ -148,10 +149,14 @@ def check_model(model, model_name, x, y, check_model_io=True): :param check_model_io: :return: ''' + early_stopping = EarlyStopping(monitor='val_acc', min_delta=0, verbose=1, patience=0, mode='max') + model_checkpoint = ModelCheckpoint(filepath='model.ckpt', monitor='val_acc', verbose=1, + save_best_only=True, + save_weights_only=False, mode='max', period=1) model.compile('adam', 'binary_crossentropy', - metrics=['binary_crossentropy']) - model.fit(x, y, batch_size=100, epochs=1, validation_split=0.5) + metrics=['binary_crossentropy', 'acc']) + model.fit(x, y, batch_size=100, epochs=1, validation_split=0.5, callbacks=[early_stopping, model_checkpoint]) print(model_name + 'test, train valid pass!') torch.save(model.state_dict(), model_name + '_weights.h5') @@ -165,9 +170,10 @@ def check_model(model, model_name, x, y, check_model_io=True): print(model_name + 'test save load model pass!') print(model_name + 'test pass!') -def get_device(use_cuda = True): + +def get_device(use_cuda=True): device = 'cpu' if use_cuda and torch.cuda.is_available(): print('cuda ready...') device = 'cuda:0' - return device \ No newline at end of file + return device