diff --git a/examples/text/README.md b/examples/text/README.md
index 242dbfe..c027968 100644
--- a/examples/text/README.md
+++ b/examples/text/README.md
@@ -15,19 +15,20 @@ conda activate discrete_flow_matching
 
 ## Usage
 
-To train a discrete flow matching model on fine-web-edu, run:
-
+Specify the data cache and checkpoint directories. Data will automatically be downloaded into the cache directory.
 ```bash
 CACHE_DIR=...
+HYDRA_RUN_DIR=...
+```
+
+To train a discrete flow matching model on fine-web-edu, run:
 
+```bash
 python run_train.py data.cache_dir=${CACHE_DIR}
 ```
 
 To use `slurm`, modify the `slurm` config according to the cluster you are working on, and run:
 ```bash
-CACHE_DIR=...
-HYDRA_RUN_DIR=...
-
 python run_train.py data.cache_dir=${CACHE_DIR} hydra_dir=${HYDRA_RUN_DIR} -m &
 ```
 
@@ -119,4 +120,4 @@ This example partially use code from:
 
 The majority of the code in this example is licensed under CC-BY-NC, however portions of the project are available under separate license terms: 
 - flash attention and TorchData are under BSD 3 license.
-- Discrete Diffusion Modeling by Estimating the Ratios of the Data Distribution and GLIDE: Towards Photorealistic Image Generation and Editing with Text-Guided Diffusion Models are under MIT license.
\ No newline at end of file
+- Discrete Diffusion Modeling by Estimating the Ratios of the Data Distribution and GLIDE: Towards Photorealistic Image Generation and Editing with Text-Guided Diffusion Models are under MIT license.