Merge branch 'master' of https://github.com/zhengbi-yong/PPG

COMMANDS.md

@@ -1,5 +1,12 @@
 # 常用命令
-
+## 下载数据集
+```zsh
+python extracode/download_dataset/PPG_FieldStudy.py
+```
+## 可视化数据集
+```zsh
+python extracode/visualize/PPG_FieldStudy_pickle.py
+```
 ## 训练
 
 ### WINDOWS+GPU

configs/data/PPG_FieldStudy.yaml

@@ -1,5 +1,5 @@
 _target_: src.data.PPGFieldStudyDatamodule.PPGDataModule
-data_dir: "./data/PPG_FieldStudy"
+data_dir: "./data/PPG_FieldStudy/PPG_FieldStudy"
 window_size: 256
 stride: 128
 batch_size: 64

configs/data/PPG_FieldStudy_MvCFT.yaml

@@ -0,0 +1,9 @@
+_target_: src.data.PPGFieldStudyMvCFTDatamodule.PPGDataModule
+data_dir: "./data/PPG_FieldStudy/PPG_FieldStudy"
+window_size: 256
+stride: 128
+batch_size: 64
+num_workers: 4 # Set to 0 for debugging
+train_split: 0.7
+val_split: 0.15
+test_split: 0.15

configs/experiment/PPG_FieldStudy_MvCFT.yaml

@@ -0,0 +1,41 @@
+# @package _global_
+
+# to execute this experiment run:
+# python train.py experiment=example
+
+defaults:
+  - override /data: PPG_FieldStudy_MvCFT
+  - override /model: PPGMvCFT
+  - override /callbacks: default
+  - override /trainer: default
+
+# all parameters below will be merged with parameters from default configurations set above
+# this allows you to overwrite only specified parameters
+
+tags: ["PPG_FieldStudy", "PPGMvCFT"]
+
+seed: 12345
+
+trainer:
+  min_epochs: 10
+  max_epochs: 50
+  gradient_clip_val: 0.5
+  accelerator: "gpu"
+  devices: 1
+  callbacks: [progress_bar]
+  enable_checkpointing: True
+
+model:
+  input_dim: 14
+  d_model: 128
+  lr: 1e-3
+
+data:
+  batch_size: 64
+
+logger:
+  wandb:
+    tags: ${tags}
+    group: "PPG_FieldStudy"
+  aim:
+    experiment: "PPG_FieldStudy"

configs/model/PPGMvCFT.yaml

@@ -0,0 +1,5 @@
+_target_: src.models.PPGMvCFTModule.PPGMvCFTModule
+
+input_dim: 14
+d_model: 128
+lr: 1e-3

doc/prompt/model.md

@@ -0,0 +1,10 @@
+```csv
+date,segment,time_sec,time_min,frequency_hz,amplitude,phase,label
+2024-11-17,1,7.0,0.11666666666666667,37993.1,12.93741,-4.32236,0
+2024-11-17,1,7.0,0.11666666666666667,41437.9,12.87911,-4.04065,0
+2024-11-17,1,8.0,0.13333333333333333,44882.8,12.78722,-3.65708,0
+2024-11-17,1,8.0,0.13333333333333333,48327.6,12.73593,-3.31598,0
+2024-11-17,1,9.0,0.15,51772.4,12.69731,-2.83716,0
+```
+我的数据如上，请你基于Pytorch写一个血糖三分类的模型，label 0 代表低，1代表中，2代表高。
+请你将数据按照1分钟的长度来进行划分，写出dataset和dataloader，并帮我写出完整的训练验证以及测试流程代码。

doc/prompt/preprocess.md

@@ -0,0 +1,65 @@
+我现在需要做一个血糖高中低三分类的项目，我的数据集如下
+```zsh
+(base)  sisyphus@sisyphus-dual4090  ~/Projects/PPG/data   master  tree GlucoseBIT
+GlucoseBIT
+├── 2024-11-13
+│   ├── 1-24.txt
+│   ├── 2-24.txt
+│   └── 3-24.txt
+├── 2024-11-17
+│   ├── 1-10.txt
+│   ├── 2-12.txt
+│   ├── 3-10.txt
+│   ├── 4-12.txt
+│   ├── 5-10.txt
+│   └── 6-12.txt
+└── 2024-11-18
+    ├── 1-12.txt
+    ├── 2-12.txt
+    ├── 3-12.txt
+    ├── 4-10.txt
+    ├── 5-12.txt
+    ├── 6-10.txt
+    └── 7-10.txt
+
+3 directories, 16 files
+```
+每个txt文件的文件名意义如下：
+-号前面的数字代表第几段，-号后面的数字代表持续时间，
+例如2024-11-17的3-10.txt文件中记录的是当次实验的第22-32分钟时间段的数据。
+每个txt内部的数据如下
+```txt
+ 00:00:02 
+6989.7 hz  
+15.40351     -7.47591 
+ 00:00:03 
+10434.5 hz  
+15.56827     -9.31106 
+ 00:00:03 
+13879.3 hz  
+15.69426     -14.58940 
+ 00:00:04 
+17324.1 hz  
+15.27117     -19.40495 
+ 00:00:04 
+20769.0 hz  
+15.15964     -20.27624 
+ 00:00:05 
+24213.8 hz  
+14.57288     -18.74838 
+ 00:00:05 
+27658.6 hz  
+14.27791     -15.12476 
+ 00:00:06 
+31103.4 hz  
+14.10476     -9.03115 
+```
+以下面的数据单元为例
+ 00:00:06 
+31103.4 hz  
+14.10476     -9.03115 
+最上面的是时间，第二行是频率，第三行左边的是幅值，右边的是相位。
+
+请你完成几件事情：
+1.将每一天的数据集合并成一份，并且存储为csv格式，并且根据时间打上标签0-10分钟和65分钟之后是低，10-25和45-65分钟是中，25-45是高。
+2.使用Pytorch框架来帮我写一个血糖三分类的模型

extracode/download_dataset/PPG_FieldStudy.py

@@ -0,0 +1,29 @@
+import kagglehub
+import os
+import shutil
+
+# 下载数据集到默认的缓存路径
+path = kagglehub.dataset_download("dishantvishwakarma/ppg-dataset-shared")
+print("下载完成，数据集路径为:", path)
+
+# 设定目标文件夹路径，将数据集移至该脚本父目录的父目录下的 data 文件夹
+target_path = os.path.join(os.path.dirname(os.path.dirname(os.path.dirname(__file__))), "data")
+
+# 如果目标文件夹不存在，创建该文件夹
+os.makedirs(target_path, exist_ok=True)
+
+# 将数据集文件逐个复制到目标文件夹
+for item in os.listdir(path):
+    source_item = os.path.join(path, item)
+    target_item = os.path.join(target_path, item)
+
+    # 如果目标文件已存在，跳过复制
+    if os.path.exists(target_item):
+        print(f"文件已存在，跳过复制: {target_item}")
+    else:
+        if os.path.isdir(source_item):
+            shutil.copytree(source_item, target_item)
+        else:
+            shutil.copy2(source_item, target_item)
+
+print("数据集已移动到目标路径:", target_path)