Update README

Author: Shamdan17

README.md

@@ -24,7 +24,16 @@ We propose "**E**fficiency through **T**hinking **A**head" (ETA), an asynchronou
 
 ## Getting Started <a name="gettingstarted"></a>
 
-Please refer to [TRAIN_EVAL.md](docs/TRAIN_EVAL.md) for detailed instructions on how to train and evaluate the model.
+To get started with ETA:
+- Training
+  - [Download Bench2Drive Data](docs/TRAIN_EVAL.md#b2ddata)  
+  - [Setup other prerequisites](docs/TRAIN_EVAL.md#trainingsetup)
+  - [Start training](docs/TRAIN_EVAL.md#training)
+- Evaluation
+  - [Setup Bench2Drive](docs/TRAIN_EVAL.md#evalsetup)
+  - [Setup files and configs](docs/TRAIN_EVAL.md#evalfilesetup)
+  - [Download checkpoints](docs/TRAIN_EVAL.md#evalcheckpoints)
+  - [Run evaluation](docs/TRAIN_EVAL.md#runeval)
 
 ## TODO List
 - [x] ETA Training code

docs/TRAIN_EVAL.md

@@ -1,6 +1,12 @@
 ## Training Setup <a name="trainingsetup"></a>
-### Bench2Drive data
-Download the [Bench2Drive](https://github.com/Thinklab-SJTU/Bench2Drive) dataset and unzip all the directories. 
+### Bench2Drive data <a name="b2ddata"></a>
+Download the [Bench2Drive](https://github.com/Thinklab-SJTU/Bench2Drive) dataset and unzip all the directories to a single folder. The dataset should have a structure similar to the following:
+
+```
+MainFolder/
+    HazardAtSideLaneTwoWays_Town12_Route1133_Weather15/
+    ParkingCutIn_Town12_Route765_Weather11/
+```
 
 ### Libraries
 Install the [skit](https://github.com/Shamdan17/skit) toolkit for distributed sampling and training wrappers. More dependencies are listed in the `requirements.txt` file. Install them using pip:
@@ -21,7 +27,7 @@ wandb_entity: WANDBUSERNAME
 
 Furthermore, modify the Makefile to update the username to the config name you just created.
 
-### Training
+### Training <a name="training"></a>
 
 To train the base and async models on 8 nodes with 4 GPUs each, run the following commands from the `carformer` directory. We run these commands in a SLURM job. Directly using make will not properly parallelize the training across multiple nodes, so modify it accordingly if not working in a SLURM environment.
 
@@ -36,7 +42,7 @@ make ETA_async_model_s42
 
 [Bench2Drive](https://github.com/Thinklab-SJTU/Bench2Drive) is required for evaluation. Please follow the "Eval Tools" section.
 
-### File setup:
+### File setup: <a name="evalfilesetup"></a>
 
 Only follow these steps **AFTER** Bench2Drive is set up following the Bench2Drive instructions. Please place the files found in "misc" and "team_code" in the following structure in the Bench2Drive repo:
 
@@ -64,11 +70,11 @@ working_dir: /path/to/Bench2Drive
 b2d_path: /path/to/Bench2Drive
 ```
 
-### Download checkpoints:
+### Download checkpoints: <a name="evalcheckpoints"></a>
 
 Checkpoints uploading is in progress.
 
-### Running the evaluation:
+### Running the evaluation: <a name="runeval"></a>
 
 **Note:** Unlike Bench2Drive's setup, this evaluation code requires a separate instance of CARLA running. It will NOT run CARLA for you.