Upload DCRNN model trained on METR-LA

README.md

@@ -0,0 +1,63 @@
+---
+tags:
+- traffic-forecasting
+- time-series
+- graph-neural-network
+- dcrnn
+datasets:
+- metr-la
+---
+
+# DCRNN Model - METR-LA
+
+Diffusion Convolutional Recurrent Neural Network (DCRNN) trained on METR-LA dataset for traffic speed forecasting.
+
+## Model Description
+
+This model uses a graph neural network architecture that combines:
+- Diffusion convolution to capture spatial dependencies on road networks
+- Recurrent neural networks (GRU) for temporal modeling
+- Sequence-to-sequence learning for multi-step ahead forecasting
+
+## Evaluation Metrics
+
+- **test_mae**: 12.0126
+
+
+## Dataset
+
+**METR-LA**: Traffic speed data from highway sensors.
+
+## Usage
+
+```python
+from dcrnn_pytorch.model.pytorch.dcrnn_supervisor import DCRNNSupervisor
+from utils.dcrnn import load_from_hub
+
+# Download checkpoint
+checkpoint_path = load_from_hub("your-username/dcrnn-metr-la")
+
+# Load model (requires config and adj_mx from checkpoint)
+# See documentation for full loading example
+```
+
+## Training
+
+Model was trained using the DCRNN PyTorch implementation with default hyperparameters.
+
+## Citation
+
+If you use this model, please cite the original DCRNN paper:
+
+```
+@inproceedings{li2018dcrnn,
+  title={Diffusion Convolutional Recurrent Neural Network: Data-Driven Traffic Forecasting},
+  author={Li, Yaguang and Yu, Rose and Shahabi, Cyrus and Liu, Yan},
+  booktitle={International Conference on Learning Representations (ICLR)},
+  year={2018}
+}
+```
+
+## License
+
+This model checkpoint is released under the same license as the training code.

config.json

@@ -0,0 +1,49 @@
+{
+  "epoch": 0,
+  "scaler": {
+    "mean": "54.405926",
+    "std": "19.49374"
+  },
+  "model_config": {
+    "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
+  },
+  "data_config": {
+    "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"
+  },
+  "train_config": {
+    "base_lr": 0.01,
+    "dropout": 0,
+    "epoch": 0,
+    "epochs": 100,
+    "epsilon": 0.001,
+    "global_step": 0,
+    "lr_decay_ratio": 0.1,
+    "max_grad_norm": 5,
+    "max_to_keep": 100,
+    "min_learning_rate": 2e-06,
+    "optimizer": "adam",
+    "patience": 50,
+    "steps": [
+      20,
+      30,
+      40,
+      50
+    ],
+    "test_every_n_epochs": 10
+  }
+}

dcrnn_model.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d07a8af6668d280b1ecda0941b3b38199a65009efac6c8073ed6e8dbcdaf9d73
+size 1497677

metadata.json

@@ -0,0 +1,9 @@
+{
+  "dataset": "METR-LA",
+  "upload_date": "2025-11-01T21:53:26.458756",
+  "metrics": {
+    "test_mae": 12.012598443914342
+  },
+  "framework": "PyTorch",
+  "model_type": "DCRNN"
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1306206e5996a5336b7f7df832fc7a0aeb32148489262d84f07a7d7573895b92
+size 1491428