Merge pull request #9 from liyaguang/v2

Upgrading to DCRNN v2.

README.md

@@ -9,7 +9,6 @@ Yaguang Li, Rose Yu, Cyrus Shahabi, Yan Liu, [Diffusion Convolutional Recurrent
 
 
 ## Requirements
-- hyperopt>=0.1
 - scipy>=0.19.0
 - numpy>=1.12.1
 - pandas>=0.19.2
@@ -22,34 +21,38 @@ Dependency can be installed using the following command:
 pip install -r requirements.txt
 ```
 
-
-## Traffic Data
+## Data Preparation
 The traffic data file for Los Angeles, i.e., `df_highway_2012_4mon_sample.h5`, is available [here](https://drive.google.com/open?id=1tjf5aXCgUoimvADyxKqb-YUlxP8O46pb), and should be
-put into the `data/` folder.
+put into the `data/METR-LA` folder.
 Besides, the locations of sensors are available at [data/sensor_graph/graph_sensor_locations.csv](https://github.com/liyaguang/DCRNN/blob/master/data/sensor_graph/graph_sensor_locations.csv).
+```bash
+python -m scripts.generate_training_data --output_dir=data/METR-LA
+```
+The generated train/val/test dataset will be saved at `data/METR-LA/{train,val,test}.npz`.
 
-## Graph Construction
- As the currently implementation is based on pre-calculated road network distances between sensors, it currently only
- supports sensor ids in Los Angeles (see `data/sensor_graph/sensor_info_201206.csv`).
+
+## Run the Pre-trained Model
 
 ```bash
-python gen_adj_mx.py  --sensor_ids_filename=data/sensor_graph/graph_sensor_ids.txt --normalized_k=0.1\
-    --output_pkl_filename=data/sensor_graph/adj_mx.pkl
+python run_demo.py
 ```
+The generated prediction of DCRNN is in `data/results/dcrnn_predictions_[1-12].h5`.
 
-## Train the Model
+
+## Model Training
 ```bash
 python dcrnn_train.py --config_filename=data/model/dcrnn_config.yaml
 ```
+Each epoch takes about 5min with a single GTX 1080 Ti. There is a chance that train/val loss will explode,  gradient explosion, 
 
-
-## Run the Pre-trained Model
+## Graph Construction
+ As the currently implementation is based on pre-calculated road network distances between sensors, it currently only
+ supports sensor ids in Los Angeles (see `data/sensor_graph/sensor_info_201206.csv`).
 
 ```bash
-python run_demo.py
+python gen_adj_mx.py  --sensor_ids_filename=data/sensor_graph/graph_sensor_ids.txt --normalized_k=0.1\
+    --output_pkl_filename=data/sensor_graph/adj_mx.pkl
 ```
-The generated prediction of DCRNN is in `data/results/dcrnn_predictions_[1-12].h5`.
-
 
 More details are being added ...
 

data/model/dcrnn_config.yaml

@@ -1,34 +1,38 @@
 ---
 base_dir: data/model
-batch_size: 64
-cl_decay_steps: 2000
-data_type: ALL
-dropout: 0
-epoch: 0
-epochs: 100
-filter_type: dual_random_walk
-global_step: 0
-graph_pkl_filename: data/sensor_graph/adj_mx.pkl
-horizon: 12
-l1_decay: 0
-learning_rate: 0.01
-loss_func: MAE
-lr_decay: 0.1
-lr_decay_epoch: 20
-lr_decay_interval: 10
-max_diffusion_step: 2
-max_grad_norm: 5
-min_learning_rate: 2.0e-06
-null_val: 0
-num_rnn_layers: 2
-output_dim: 1
-patience: 50
-rnn_units: 64
-seq_len: 12
-test_every_n_epochs: 10
-test_ratio: 0.2
-use_cpu_only: false
-use_curriculum_learning: true
-validation_ratio: 0.1
-verbose: 0
-write_db: false
+data:
+  batch_size: 64
+  dataset_dir: data/METR-LA
+  test_batch_size: 64
+  val_batch_size: 64
+  graph_pkl_filename: data/sensor_graph/adj_mx.pkl
+
+model:
+  cl_decay_steps: 2000
+  filter_type: dual_random_walk
+  horizon: 12
+  input_dim: 2
+  l1_decay: 0
+  max_diffusion_step: 2
+  num_nodes: 207
+  num_rnn_layers: 2
+  output_dim: 1
+  rnn_units: 64
+  seq_len: 12
+  use_curriculum_learning: true
+
+train:
+  base_lr: 0.01
+  dropout: 0
+  epoch: 0
+  epochs: 100
+  epsilon: 1.0e-3
+  global_step: 0
+  lr_decay_ratio: 0.1
+  max_grad_norm: 5
+  max_to_keep: 100
+  min_learning_rate: 2.0e-06
+  optimizer: adam
+  patience: 50
+  steps: [20, 30, 40, 50]
+  test_every_n_epochs: 10

data/model/dcrnn_config_u16_lap.yaml

@@ -0,0 +1,36 @@
+---
+base_dir: data/model
+data:
+  batch_size: 64
+  dataset_dir: data/METR-LA
+  test_batch_size: 64
+  val_batch_size: 64
+  graph_pkl_filename: data/sensor_graph/adj_mx.pkl
+
+model:
+  cl_decay_steps: 2000
+  filter_type: laplacian
+  horizon: 12
+  input_dim: 2
+  l1_decay: 0
+  max_diffusion_step: 2
+  max_grad_norm: 5
+  num_nodes: 207
+  num_rnn_layers: 2
+  output_dim: 1
+  rnn_units: 16
+  seq_len: 12
+  use_curriculum_learning: true
+
+train:
+  base_lr: 0.01
+  dropout: 0
+  epoch: 0
+  epochs: 100
+  global_step: 0
+  lr_decay_ratio: 0.1
+  steps: [20, 30, 40, 50]
+  max_to_keep: 100
+  min_learning_rate: 2.0e-06
+  patience: 50
+  test_every_n_epochs: 10

data/model/dcrnn_test_config.yaml

@@ -1,35 +1,36 @@
 ---
 base_dir: data/model
-batch_size: 64
-cl_decay_steps: 2000
-data_type: ALL
-dropout: 0
-epoch: 0
-epochs: 100
-filter_type: random_walk
-global_step: 0
-graph_pkl_filename: data/sensor_graph/adj_mx.pkl
-horizon: 3
-l1_decay: 0
-learning_rate: 0.01
-loss_func: MAE
-lr_decay: 0.1
-lr_decay_epoch: 20
-lr_decay_interval: 10
-max_diffusion_step: 2
-max_grad_norm: 5
-method_type: GCRNN
-min_learning_rate: 2.0e-06
-null_val: 0
-num_rnn_layers: 2
-output_dim: 1
-patience: 50
-rnn_units: 16
-seq_len: 3
-test_every_n_epochs: 10
-test_ratio: 0.2
-use_cpu_only: false
-use_curriculum_learning: true
-validation_ratio: 0.1
-verbose: 0
-write_db: false
+data:
+  batch_size: 64
+  dataset_dir: data/METR-LA
+  test_batch_size: 64
+  val_batch_size: 64
+  graph_pkl_filename: data/sensor_graph/adj_mx.pkl
+
+model:
+  cl_decay_steps: 2000
+  filter_type: dual_random_walk
+  horizon: 12
+  input_dim: 2
+  l1_decay: 0
+  max_diffusion_step: 2
+  max_grad_norm: 5
+  num_nodes: 207
+  num_rnn_layers: 2
+  output_dim: 1
+  rnn_units: 64
+  seq_len: 12
+  use_curriculum_learning: true
+
+train:
+  base_lr: 0.01
+  dropout: 0
+  epoch: 0
+  epochs: 100
+  global_step: 0
+  lr_decay_ratio: 0.1
+  steps: [20, 30, 40, 50]
+  max_to_keep: 100
+  min_learning_rate: 2.0e-06
+  patience: 50
+  test_every_n_epochs: 10

data/model/pretrained/config.yaml

@@ -0,0 +1,40 @@
+base_dir: data/model
+data:
+  batch_size: 64
+  dataset_dir: data/METR-LA
+  graph_pkl_filename: data/sensor_graph/adj_mx.pkl
+  test_batch_size: 64
+model:
+  cl_decay_steps: 2000
+  filter_type: dual_random_walk
+  horizon: 12
+  input_dim: 2
+  l1_decay: 0
+  max_diffusion_step: 2
+  num_nodes: 207
+  num_rnn_layers: 2
+  output_dim: 1
+  rnn_units: 64
+  seq_len: 12
+  use_curriculum_learning: true
+train:
+  base_lr: 0.01
+  dropout: 0
+  epoch: 64
+  epochs: 100
+  epsilon: 0.001
+  global_step: 24375
+  log_dir: data/model/pretrained/
+  lr_decay_ratio: 0.1
+  max_grad_norm: 5
+  max_to_keep: 100
+  min_learning_rate: 2.0e-06
+  model_filename: data/model/pretrained/models-2.7422-24375
+  optimizer: adam
+  patience: 50
+  steps:
+  - 20
+  - 30
+  - 40
+  - 50
+  test_every_n_epochs: 10