Skip to content

P
PaddleRec

PaddlePaddle / PaddleRec

通知 68
Star 12
Fork 5
P
PaddleRec
提交 887fdc4f 编写于 作者: Y yinhaofeng
浏览文件
操作

readme

上级 ccbe52fd
显示空白变更内容
内联 并排
Showing with 28 addition and 23 deletion
models/rank/logistic_regression/config.yaml
浏览文件 @ 887fdc4f
...@@ -55,11 +55,13 @@ runner: ...@@ -55,11 +55,13 @@ runner:
save_checkpoint_path: "increment" save_checkpoint_path: "increment"
save_inference_path: "inference" save_inference_path: "inference"
print_interval: 1 print_interval: 1
phases: phase1
- name: infer_runner - name: infer_runner
class: infer class: infer
device: cpu device: cpu
init_model_path: "increment/1" init_model_path: "increment/1"
print_interval: 1 print_interval: 1
phases: infer_phase
phase: phase:
......
models/rank/logistic_regression/readme.md
浏览文件 @ 887fdc4f
...@@ -112,7 +112,7 @@ logistic_regression模型的组网比较直观，本质是一个二分类任务 ...@@ -112,7 +112,7 @@ logistic_regression模型的组网比较直观，本质是一个二分类任务
### sigmoid层 ### sigmoid层
将离散数据通过embedding查表得到的值，与连续数据的输入进行相乘再累加的操作，合为一个整体输入。我们又构造了一个初始化为0，shape为1的变量，将其与累加结果相加一起输入sigmoid中得到分类结果。 将离散数据通过embedding查表得到的值，与连续数据的输入进行相乘再累加的操作，合为一个整体输入。我们又构造了一个初始化为0，shape为1的变量，将其与累加结果相加一起输入sigmoid中得到分类结果。
在这里，可以将这个过程理解为一个全连接层。通过embedding查表获得权重w，构造的变量b_linear即为偏置变量b，再经过激活函数为sigmoid。 在这里，可以将这个过程理解为一个全连接层。通过embedding查表获得权重w，构造的变量b_linear即为偏置变量b，再经过激活函数为sigmoid得到输出。
### Loss及Auc计算 ### Loss及Auc计算
- 预测的结果通过直接通过激活函数sigmoid给出，为了得到每条样本分属于正负样本的概率，我们将预测结果和`1-predict`合并起来得到predict_2d，以便接下来计算auc。 - 预测的结果通过直接通过激活函数sigmoid给出，为了得到每条样本分属于正负样本的概率，我们将预测结果和`1-predict`合并起来得到predict_2d，以便接下来计算auc。
...@@ -128,7 +128,7 @@ logistic_regression模型的组网比较直观，本质是一个二分类任务 ...@@ -128,7 +128,7 @@ logistic_regression模型的组网比较直观，本质是一个二分类任务
| 模型 | auc | batch_size | thread_num| epoch_num| Time of each epoch | | 模型 | auc | batch_size | thread_num| epoch_num| Time of each epoch |
| :------| :------ | :------| :------ | :------| :------ | | :------| :------ | :------| :------ | :------| :------ |
| LR | 0.7243 | 1024 | 10 | 2 | 约3小时 | | LR | 0.7611 | 1024 | 10 | 2 | 约4小时 |
1. 确认您当前所在目录为PaddleRec/models/rank/deepfm 1. 确认您当前所在目录为PaddleRec/models/rank/deepfm
2. 在data目录下运行数据一键处理脚本，命令如下： 2. 在data目录下运行数据一键处理脚本，命令如下：
...@@ -143,6 +143,7 @@ cd .. ...@@ -143,6 +143,7 @@ cd ..
将train_sample中的data_path改为{workspace}/data/slot_train_data 将train_sample中的data_path改为{workspace}/data/slot_train_data
将infer_sample中的batch_size从5改为1024 将infer_sample中的batch_size从5改为1024
将infer_sample中的data_path改为{workspace}/data/slot_test_data 将infer_sample中的data_path改为{workspace}/data/slot_test_data
根据自己的需求调整phase中的线程数
4. 运行命令，模型会进行两个epoch的训练，然后预测第二个epoch，并获得相应auc指标 4. 运行命令，模型会进行两个epoch的训练，然后预测第二个epoch，并获得相应auc指标
``` ```
python -m paddlerec.run -m ./config.yaml python -m paddlerec.run -m ./config.yaml
...@@ -158,28 +159,30 @@ Warning:please make sure there are no hidden files in the dataset folder and che ...@@ -158,28 +159,30 @@ Warning:please make sure there are no hidden files in the dataset folder and che
Warning:please make sure there are no hidden files in the dataset folder and check these hidden files:[] Warning:please make sure there are no hidden files in the dataset folder and check these hidden files:[]
Running SingleInferStartup. Running SingleInferStartup.
Running SingleInferRunner. Running SingleInferRunner.
load persistables from increment/0 load persistables from increment/1
2020-09-18 11:43:23,533-INFO: [Infer] batch: 1, time_each_interval: 0.18s, AUC: [0.72274697] 2020-09-25 01:38:01,653-INFO: [Infer] batch: 1, time_each_interval: 0.64s, AUC: [0.76076558]
2020-09-18 11:43:23,564-INFO: [Infer] batch: 2, time_each_interval: 0.03s, AUC: [0.72274716] 2020-09-25 01:38:01,890-INFO: [Infer] batch: 2, time_each_interval: 0.24s, AUC: [0.76076588]
2020-09-18 11:43:23,624-INFO: [Infer] batch: 3, time_each_interval: 0.06s, AUC: [0.72274746] 2020-09-25 01:38:02,116-INFO: [Infer] batch: 3, time_each_interval: 0.23s, AUC: [0.76076599]
2020-09-18 11:43:23,695-INFO: [Infer] batch: 4, time_each_interval: 0.07s, AUC: [0.72274772] 2020-09-25 01:38:02,351-INFO: [Infer] batch: 4, time_each_interval: 0.23s, AUC: [0.76076598]
2020-09-18 11:43:23,841-INFO: [Infer] batch: 5, time_each_interval: 0.15s, AUC: [0.72274817] 2020-09-25 01:38:02,603-INFO: [Infer] batch: 5, time_each_interval: 0.25s, AUC: [0.76076637]
2020-09-18 11:43:23,922-INFO: [Infer] batch: 6, time_each_interval: 0.08s, AUC: [0.72274794] 2020-09-25 01:38:02,841-INFO: [Infer] batch: 6, time_each_interval: 0.24s, AUC: [0.76076656]
2020-09-18 11:43:23,989-INFO: [Infer] batch: 7, time_each_interval: 0.07s, AUC: [0.72274796] 2020-09-25 01:38:03,076-INFO: [Infer] batch: 7, time_each_interval: 0.24s, AUC: [0.76076668]
2020-09-18 11:43:24,058-INFO: [Infer] batch: 8, time_each_interval: 0.07s, AUC: [0.72274792] 2020-09-25 01:38:03,308-INFO: [Infer] batch: 8, time_each_interval: 0.23s, AUC: [0.76076662]
2020-09-18 11:43:24,130-INFO: [Infer] batch: 9, time_each_interval: 0.07s, AUC: [0.72274824] 2020-09-25 01:38:03,541-INFO: [Infer] batch: 9, time_each_interval: 0.23s, AUC: [0.76076698]
2020-09-18 11:43:24,195-INFO: [Infer] batch: 10, time_each_interval: 0.07s, AUC: [0.72274831] 2020-09-25 01:38:03,772-INFO: [Infer] batch: 10, time_each_interval: 0.23s, AUC: [0.76076676]
2020-09-25 01:38:04,025-INFO: [Infer] batch: 11, time_each_interval: 0.25s, AUC: [0.76076655]
... ...
2020-09-18 12:57:53,777-INFO: [Infer] batch: 17959, time_each_interval: 0.07s, AUC: [0.72434065] 2020-09-25 02:00:14,043-INFO: [Infer] batch: 4482, time_each_interval: 0.26s, AUC: [0.76117275]
2020-09-18 12:57:53,848-INFO: [Infer] batch: 17960, time_each_interval: 0.07s, AUC: [0.72434041] 2020-09-25 02:00:14,338-INFO: [Infer] batch: 4483, time_each_interval: 0.29s, AUC: [0.7611731]
2020-09-18 12:57:53,910-INFO: [Infer] batch: 17961, time_each_interval: 0.06s, AUC: [0.72434046] 2020-09-25 02:00:14,614-INFO: [Infer] batch: 4484, time_each_interval: 0.27s, AUC: [0.76117289]
2020-09-18 12:57:53,974-INFO: [Infer] batch: 17962, time_each_interval: 0.06s, AUC: [0.72434055] 2020-09-25 02:00:14,858-INFO: [Infer] batch: 4485, time_each_interval: 0.25s, AUC: [0.76117328]
2020-09-18 12:57:54,045-INFO: [Infer] batch: 17963, time_each_interval: 0.07s, AUC: [0.72434008] 2020-09-25 02:00:15,187-INFO: [Infer] batch: 4486, time_each_interval: 0.33s, AUC: [0.7611733]
2020-09-18 12:57:54,111-INFO: [Infer] batch: 17964, time_each_interval: 0.07s, AUC: [0.72434022] 2020-09-25 02:00:15,483-INFO: [Infer] batch: 4487, time_each_interval: 0.30s, AUC: [0.76117372]
2020-09-18 12:57:54,177-INFO: [Infer] batch: 17965, time_each_interval: 0.07s, AUC: [0.72434011] 2020-09-25 02:00:15,729-INFO: [Infer] batch: 4488, time_each_interval: 0.25s, AUC: [0.76117397]
2020-09-18 12:57:54,246-INFO: [Infer] batch: 17966, time_each_interval: 0.07s, AUC: [0.72434023] 2020-09-25 02:00:15,965-INFO: [Infer] batch: 4489, time_each_interval: 0.24s, AUC: [0.7611739]
2020-09-18 12:57:54,309-INFO: [Infer] batch: 17967, time_each_interval: 0.06s, AUC: [0.72434046] 2020-09-25 02:00:16,244-INFO: [Infer] batch: 4490, time_each_interval: 0.28s, AUC: [0.76117379]
Infer infer_phase of epoch increment/0 done, use time: 1414.92181587, global metrics: AUC=0.72434046 2020-09-25 02:00:16,560-INFO: [Infer] batch: 4491, time_each_interval: 0.32s, AUC: [0.76117405]
Infer infer_phase of epoch increment/1 done, use time: 1335.62605906, global metrics: AUC=0.76117405
PaddleRec Finish PaddleRec Finish
``` ```
......
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或