diff --git a/README.md b/README.md
index 59887cc..57f0ede 100644
--- a/README.md
+++ b/README.md
@@ -22,7 +22,7 @@ https://github.com/TencentARC/InstantMesh/assets/20635237/dab3511e-e7c6-4c0b-bab
 - [x] Release inference and training code.
 - [x] Release model weights.
 - [x] Release huggingface gradio demo. Please try it at [demo](https://huggingface.co/spaces/TencentARC/InstantMesh) link.
-- [ ] Add support to low-memory GPU environment.
+- [x] Add support to low-memory GPU environment.
 - [ ] Add support to more multi-view diffusion models.
 
 # ⚙️ Dependencies and Installation
diff --git a/app.py b/app.py
index 032f112..c74140d 100644
--- a/app.py
+++ b/app.py
@@ -86,6 +86,7 @@ pipeline = DiffusionPipeline.from_pretrained(
     "sudo-ai/zero123plus-v1.2", 
     custom_pipeline="zero123plus",
     torch_dtype=torch.float16,
+    cache_dir=model_cache_dir
 )
 pipeline.scheduler = EulerAncestralDiscreteScheduler.from_config(
     pipeline.scheduler.config, timestep_spacing='trailing'
diff --git a/docker/README.md b/docker/README.md
index 3062135..68646d0 100644
--- a/docker/README.md
+++ b/docker/README.md
@@ -1,13 +1,22 @@
 # Docker setup
 
-This docker setup is tested on WSL(Ubuntu).
+This docker setup is tested on Ubuntu.
 
 make sure you are under directory yourworkspace/instantmesh/
 
-run
+Build docker image:
 
-`docker build -t instantmesh/deploy:cuda12.1 -f docker/Dockerfile .`
+```bash
+docker build -t instantmesh -f docker/Dockerfile .
+```
 
-then run
+Run docker image with a local model cache (so it is fast when container is started next time):
 
-`docker run --gpus all -it instantmesh/deploy:cuda12.1`
\ No newline at end of file
+```bash
+mkdir -p $HOME/models/
+export MODEL_DIR=$HOME/models/
+
+docker run -it -p 43839:43839 --platform=linux/amd64 --gpus all -v $MODEL_DIR:/workspace/instantmesh/models instantmesh
+```
+
+Navigate to `http://localhost:43839` to use the demo.