[Example] Add TGCN Model for Traffic Forecasting #972

casia-rxwang · 2024-08-15T08:38:02Z

PR types

Others

PR changes

Others

Describe

add TGCN model for traffic forecasting
add docs for TGCN
add example for TGCN
add interface for PEMSD4 & PEMSD8 dataset

* add TGCN docs * add TGCN model * add TGCN example * add PEMSD4 & PEMSD8 dataset

paddle-bot · 2024-08-15T08:38:07Z

Thanks for your contribution!

CLAassistant · 2024-08-15T08:38:08Z

All committers have signed the CLA.

HydrogenSulfate · 2024-08-16T08:57:27Z

docs/zh/examples/tgcn.md

+    # Train
+    python PaddleScience/examples/tgcn/run.py data_name=PEMSD8
+    # python PaddleScience/examples/tgcn/run.py data_name=PEMSD4


案例的默认执行路径在对应的案例文件夹下，而不是PaddleScience同级的目录下

Suggested change

# Train

python PaddleScience/examples/tgcn/run.py data_name=PEMSD8

# python PaddleScience/examples/tgcn/run.py data_name=PEMSD4

python run.py data_name=PEMSD8

# python run.py data_name=PEMSD4

HydrogenSulfate · 2024-08-16T08:57:45Z

docs/zh/examples/tgcn.md

+    # Eval
+    python PaddleScience/examples/tgcn/run.py data_name=PEMSD8 mode=eval
+    # python PaddleScience/examples/tgcn/run.py data_name=PEMSD4 mode=eval


案例的默认执行路径在对应的案例文件夹下，而不是PaddleScience同级的目录下

Suggested change

# Eval

python PaddleScience/examples/tgcn/run.py data_name=PEMSD8 mode=eval

# python PaddleScience/examples/tgcn/run.py data_name=PEMSD4 mode=eval

python run.py data_name=PEMSD8 mode=eval

# python run.py data_name=PEMSD4 mode=eval

HydrogenSulfate · 2024-08-16T09:01:30Z

docs/zh/examples/tgcn.md

+开始训练、评估前，请下载数据集：[PEMSD4 & PEMSD8](https://paddle-org.bj.bcebos.com/paddlescience/datasets/tgcn/tgcn_data.zip)。将解压后的数据集文件夹与 `PaddleScience` 文件夹放置于同一目录下。
+
+开始评估前，请下载或训练生成预训练模型：[PEMSD4](https://paddle-org.bj.bcebos.com/paddlescience/models/tgcn/PEMSD4_pretrained_model.pdparams) & [PEMSD8](https://paddle-org.bj.bcebos.com/paddlescience/models/tgcn/PEMSD8_pretrained_model.pdparams)。将预训练模型文件与 `PaddleScience` 文件夹放置于同一目录下。
+


3-5行的说明可以删除，训练和评估开始前的数据集下载命令请直接放到文档的命令中，参考如下

HydrogenSulfate · 2024-08-16T09:02:47Z

docs/zh/examples/tgcn.md

+下表展示了 TGCN 在 PEMSD4 和 PEMSD8 两个数据集上的评估结果。
+
+| 数据集 | MAE   | RMSE  |
+| :----- | :---- | :---- |
+| PEMSD4 | 21.48 | 34.06 |
+| PEMSD8 | 15.57 | 24.52 |
+


文档开头位置加上数据集和对应的预训练模型，如下所示

HydrogenSulfate · 2024-08-16T09:04:50Z

examples/tgcn/conf/run.yaml

+hydra:
+  run:
+    # dynamic output directory according to running time and override name
+    dir: __exp__/${data_name}/${now:%Y_%m_%d_%H_%M_%S}
+  job:
+    name: ${mode} # name of logfile
+    chdir: false # keep current working directory unchanged
+    config:
+      override_dirname:
+        exclude_keys:
+          - mode
+          - output_dir
+          - log_freq
+  sweep:
+    # output directory for multirun
+    dir: ${hydra.run.dir}
+    subdir: ./


Suggested change

hydra:

run:

# dynamic output directory according to running time and override name

dir: __exp__/${data_name}/${now:%Y_%m_%d_%H_%M_%S}

job:

name: ${mode} # name of logfile

chdir: false # keep current working directory unchanged

config:

override_dirname:

exclude_keys:

- mode

- output_dir

- log_freq

sweep:

# output directory for multirun

dir: ${hydra.run.dir}

subdir: ./

defaults:

- ppsci_default

- TRAIN: train_default

- TRAIN/ema: ema_default

- TRAIN/swa: swa_default

- EVAL: eval_default

- INFER: infer_default

- _self_

hydra:

run:

# dynamic output directory according to running time and override name

dir: outputs_tgcn/${now:%Y-%m-%d}/${now:%H-%M-%S}

job:

name: ${mode} # name of logfile

chdir: false # keep current working directory unchanged

callbacks:

init_callback:

_target_: ppsci.utils.callbacks.InitCallback

sweep:

# output directory for multirun

dir: ${hydra.run.dir}

subdir: ./

HydrogenSulfate · 2024-08-16T09:07:31Z

examples/tgcn/run.py

+            'input_keys': cfg.MODEL.afno.input_keys,
+            'label_keys': cfg.MODEL.afno.label_keys,


全局搜索修改一下：cfg.MODEL.afno. --> cfg.MODEL.

HydrogenSulfate · 2024-08-16T09:09:31Z

examples/tgcn/run.py

+    # set random seed for reproducibility
+    ppsci.utils.misc.set_random_seed(cfg.seed)
+
+    # initialize logger
+    logger.init_logger('ppsci', os.path.join(cfg.output_dir, 'test.log'), 'info')
+    logger.message(cfg)
+
+    # set eval dataloader config


Suggested change

# set random seed for reproducibility

ppsci.utils.misc.set_random_seed(cfg.seed)

# initialize logger

logger.init_logger('ppsci', os.path.join(cfg.output_dir, 'test.log'), 'info')

logger.message(cfg)

# set eval dataloader config

# set eval dataloader config

HydrogenSulfate · 2024-08-16T09:09:47Z

examples/tgcn/run.py

+    # set random seed for reproducibility
+    ppsci.utils.misc.set_random_seed(cfg.seed)
+
+    # initialize logger
+    logger.init_logger('ppsci', os.path.join(cfg.output_dir, 'train.log'), 'info')
+    logger.message(cfg)
+
+    # set train dataloader config


Suggested change

# set random seed for reproducibility

ppsci.utils.misc.set_random_seed(cfg.seed)

# initialize logger

logger.init_logger('ppsci', os.path.join(cfg.output_dir, 'train.log'), 'info')

logger.message(cfg)

# set train dataloader config

# set train dataloader config

HydrogenSulfate · 2024-08-16T09:12:32Z

ppsci/arch/tgcn.py

+        self.edge_index = pp.to_tensor(data=edge_index, place=cfg.device)
+        self.edge_attr = pp.to_tensor(data=edge_attr, place=cfg.device)


这两个属性好像并没有被用到？

HydrogenSulfate · 2024-08-16T09:17:07Z

ppsci/arch/tgcn.py

+
+        self.edge_index = pp.to_tensor(data=edge_index, place=cfg.device)
+        self.edge_attr = pp.to_tensor(data=edge_attr, place=cfg.device)
+        self.adj = pp.to_tensor(data=adj, place=cfg.device)


跟pytorch不同，to_tensor的place不需要手动指定，会默认使用当前全局设备设置

在__init__中生成的张量，如果不需要进行梯度更新，则建议使用如下方式注册为 buffer，

如需要梯度更新，则注册为parameter，

否则模型保存的时候，不会将这些变量持久化保存到参数文件中

HydrogenSulfate · 2024-08-16T09:18:20Z

@casia-rxwang 签署一下CLA协议：

HydrogenSulfate · 2024-08-21T05:09:52Z

@casia-rxwang 修改代码的时候顺便合并一下最新develop分支的代码，并解决一下冲突

[Feature&Doc] Add TGCN

201586a

* add TGCN docs * add TGCN model * add TGCN example * add PEMSD4 & PEMSD8 dataset

paddle-bot bot added the contributor label Aug 15, 2024

HydrogenSulfate requested changes Aug 16, 2024

View reviewed changes

luotao1 self-assigned this Aug 19, 2024

HydrogenSulfate changed the title ~~Add TGCN Model for Traffic Forecasting~~ [Example] Add TGCN Model for Traffic Forecasting Aug 21, 2024

Merge remote-tracking branch 'upstream/develop' into develop

53d9771

casia-rxwang force-pushed the develop branch 2 times, most recently from 1129e3d to 527d6a0 Compare October 9, 2024 05:54

rePR

42bf515

casia-rxwang force-pushed the develop branch from 527d6a0 to 42bf515 Compare October 9, 2024 06:28

Merge remote-tracking branch 'upstream/develop' into develop

a702fff

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[Example] Add TGCN Model for Traffic Forecasting #972

[Example] Add TGCN Model for Traffic Forecasting #972

casia-rxwang commented Aug 15, 2024

paddle-bot bot commented Aug 15, 2024

CLAassistant commented Aug 15, 2024 •

edited

Loading

HydrogenSulfate Aug 16, 2024

HydrogenSulfate Aug 16, 2024

HydrogenSulfate Aug 16, 2024

HydrogenSulfate Aug 16, 2024

HydrogenSulfate Aug 16, 2024

HydrogenSulfate Aug 16, 2024

HydrogenSulfate Aug 16, 2024

HydrogenSulfate Aug 16, 2024

HydrogenSulfate Aug 16, 2024

HydrogenSulfate Aug 16, 2024

HydrogenSulfate commented Aug 16, 2024

HydrogenSulfate commented Aug 21, 2024

		开始训练、评估前，请下载数据集：[PEMSD4 & PEMSD8](https://paddle-org.bj.bcebos.com/paddlescience/datasets/tgcn/tgcn_data.zip)。将解压后的数据集文件夹与 `PaddleScience` 文件夹放置于同一目录下。

		开始评估前，请下载或训练生成预训练模型：[PEMSD4](https://paddle-org.bj.bcebos.com/paddlescience/models/tgcn/PEMSD4_pretrained_model.pdparams) & [PEMSD8](https://paddle-org.bj.bcebos.com/paddlescience/models/tgcn/PEMSD8_pretrained_model.pdparams)。将预训练模型文件与 `PaddleScience` 文件夹放置于同一目录下。

		'input_keys': cfg.MODEL.afno.input_keys,
		'label_keys': cfg.MODEL.afno.label_keys,

		self.edge_index = pp.to_tensor(data=edge_index, place=cfg.device)
		self.edge_attr = pp.to_tensor(data=edge_attr, place=cfg.device)

[Example] Add TGCN Model for Traffic Forecasting #972

Are you sure you want to change the base?

[Example] Add TGCN Model for Traffic Forecasting #972

Conversation

casia-rxwang commented Aug 15, 2024

PR types

PR changes

Describe

paddle-bot bot commented Aug 15, 2024

CLAassistant commented Aug 15, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

HydrogenSulfate commented Aug 16, 2024

HydrogenSulfate commented Aug 21, 2024

CLAassistant commented Aug 15, 2024 •

edited

Loading