Merge 5537ea44ba into d92490b808

How_to_Run.md

@@ -0,0 +1,17 @@
+Quick Run:(All based on METR-LA)
+1. pip install -r requirements.txt
+2. Download the data from https://drive.google.com/open?id=10FOTa6HXPqX8Pf5WRoRwcFnW9BrNZEIX
+3. Put the data into data/ for making the training data
+4. Create data directories
+>> mkdir -p data/{METR-LA,PEMS-BAY}
+5. generate train/test/val dataset at data/{METR-LA,PEMS-BAY}/{train,val,test}.npz
+>> python -m scripts.generate_training_data --output_dir=data/METR-LA --traffic_df_filename=data/metr-la.h5
+6. Constructing the Graph
+>> python -m scripts.gen_adj_mx  --sensor_ids_filename=data/sensor_graph/graph_sensor_ids.txt --normalized_k=0.1\
+    --output_pkl_filename=data/sensor_graph/adj_mx.pkl
+7. Run the pre-trained model:
+>> python run_demo_pytorch.py --config_filename=data/model/pretrained/METR-LA/config.yaml
+8. Model Training 
+>> python dcrnn_train_pytorch.py --config_filename=data/model/dcrnn_la.yaml
+9. Evaluating the baselines:
+>> python -m scripts.eval_baseline_methods --traffic_reading_filename=data/metr-la.h5

data/model/pretrained/METR-LA/config.yaml

@@ -21,7 +21,7 @@ model:
 train:
   base_lr: 0.01
   dropout: 0
-  epoch: 64
+  epoch: 0
   epochs: 100
   epsilon: 0.001
   global_step: 24375

run_demo_pytorch.py

@@ -10,7 +10,7 @@ from model.pytorch.dcrnn_supervisor import DCRNNSupervisor
 
 def run_dcrnn(args):
     with open(args.config_filename) as f:
-        supervisor_config = yaml.load(f)
+        supervisor_config = yaml.load(f, Loader=yaml.Loader)
 
         graph_pkl_filename = supervisor_config['data'].get('graph_pkl_filename')
         sensor_ids, sensor_id_to_ind, adj_mx = load_graph_data(graph_pkl_filename)