DeepMatch
DeepMatch copied to clipboard
A deep matching model library for recommendations & advertising. It's easy to train models and to export representation vectors which can be used for ANN search.
Main Change: Add comirec Model based on Mind. Comirec has two version, self-attentive and dynamic routing. The dynamic routing is almost same as Mind.
作者在前N-1次迭代将变量停止了梯度,只在第N次迭代时允许梯度反向传播。请问这样实现的目的是什么,我这边尝试完全开放三次迭代的梯度,收敛效果会更好一点
感兴趣的同学关注我的个人公众号 **浅梦学习笔记** 回复 **加群** 加入我们的交流群~ 或者添加我的微信deepctrbot,备注加入DeepMatch交流群 |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)| |:--:|:--:|:--:| | [](https://github.com/shenweichen/AlgoNotes)| [](https://github.com/shenweichen/AlgoNotes)|[](https://t.zsxq.com/026UJEuzv)|
请看这一行, https://github.com/shenweichen/DeepMatch/blob/106bb4ebb3fd7c7efeb631f6de0e4d627edd0771/examples/run_youtubednn.py#L106 根据前面代码,test_true_label[uid] 应该是一个含有单值的 list(例如 [1]), 而 pred 是一个 list 列表(例如 [1,2,3,4]),但结果居然是 in 操作在一定情况下返回 true, 造成 hit 累计 1,从而最终 hit rate 不为 0。 非常古怪
当我用vocabulary_path的方式做hash,如下: VarLenSparseFeat(SparseFeat('short_item', vocabulary_size=100, embedding_dim=8,use_hash=True,vocabulary_path=index,dtype='int32',embeddings_initializer='uniform', embedding_name='item'), 4,'mean', length_name="short_sess_length") 就会有报错: File "/root/python_proj/song_matching/models/utils.py", line 342, in check_model model.fit(x, y, batch_size=10, epochs=2, validation_split=0.5) File "/root/anaconda3/envs/tf2/lib/python3.7/site-packages/tensorflow/python/keras/engine/training_v1.py", line 809, in fit use_multiprocessing=use_multiprocessing) File "/root/anaconda3/envs/tf2/lib/python3.7/site-packages/tensorflow/python/keras/engine/training_arrays.py", line 666,...
**Describe the bug(问题描述)** hi,这是一个很棒的项目,我在 youtubednn user tower 部署完成后,报错 ```shell (Pdb) stub.Predict(request) *** grpc._channel._InactiveRpcError:
Bumps [tensorflow](https://github.com/tensorflow/tensorflow) from 2.6.2 to 2.7.2. Release notes Sourced from tensorflow's releases. TensorFlow 2.7.2 Release 2.7.2 This releases introduces several vulnerability fixes: Fixes a code injection in saved_model_cli (CVE-2022-29216) Fixes...
python3.6.5 tensorflow 2.2 想把模型保存为save model格式,dssm模型成功保存,youtubematch、sdm、mind模型报错 user_embedding_model = Model(inputs=model.user_input, outputs=model.user_embedding) item_embedding_model = Model(inputs=model.item_input, outputs=model.item_embedding) # 用keras.models.save_model保存模型 from tensorflow import keras **keras.models.save_model(user_embedding_model,"./models")** 报错如下: \Python36\site-packages\tensorflow\python\keras\saving\saved_model\save_impl.py", line 566, in call_and_return_conditional_losses return layer_call(inputs, *args, **kwargs),...
util.py的get_item_embedding方法报以上异常,请问怎么支持解? 除了降级到Tf1.0还有其它办法吗?