Merge pull request #42 from kumo-ai/yiweny-patch-2

Update README.md

README.md

@@ -1,26 +1,24 @@
-# contextgnn
-
-- [Overleaf Latex link](https://www.overleaf.com/8255131161fxgzwccqftmz#5676c1)
-
-- [contextgnn blogpost link](https://docs.google.com/document/d/1kcGl9zk_pHuZ5xE9HBBVCmJa6iLiZ_yOjX9eFPpHiXw/edit)
-
-- [Spreadsheet of results](https://docs.google.com/spreadsheets/d/1bnNurVKLCgWjgvd9fCO-NexCgU75Xql9erfn6h3Wooo/edit?usp=sharing)
-
-
 ## How to Run
 
-Run [`benchmark/relbench_link_prediction_benchmark.py`](https://github.com/kumo-ai/contextgnn/blob/master/benchmark/relbench_link_prediction_benchmark.py)
+We run our experiments on NVIDIA L40S Tensor Core GPU with 44.7 GB of memory.
+If you want to run with smaller GPU memory, please set `num_layers=2` in all the scripts, unless you are running on `rel-trial` for RelBench, in this case, please use `num_layers=4`.
+
+To reproduce results on RelBench, run [`benchmark/relbench_link_prediction_benchmark.py`](https://github.com/kumo-ai/contextgnn/blob/master/benchmark/relbench_link_prediction_benchmark.py)
 
 ```sh
-python relbench_link_prediction_benchmark.py --dataset rel-trial --task site-sponsor-run --model contextgnn
+python relbench_link_prediction_benchmark.py --model contextgnn
 ```
 
+To reproduce results on IJCAI-Contest, run [`benchmark/tgt_ijcai_benchmark.py`](https://github.com/kumo-ai/ContextGNN/blob/master/benchmark/tgt_ijcai_benchmark.py)
+
+```
+python tgt_ijcai_benchmark.py --model contextgnn
+```
 
-Run [`examples/relbench_example.py`](https://github.com/kumo-ai/contextgnn/blob/master/examples/relbench_example.py)
+To run ContextGNN without optuna tuning, run
 
 ```sh
 python relbench_example.py --dataset rel-trial --task site-sponsor-run --model contextgnn
-python relbench_example.py --dataset rel-trial --task condition-sponsor-run --model contextgnn
 ```