Update README.md

README.md

@@ -8,44 +8,8 @@ tags:
 - 3d-medical-imaging
 ---
 
-# 3D Medical Diffusion VQ-AE Checkpoint
-
-This is the pretrained VQ-AE (Vector Quantized Autoencoder) checkpoint from [3D-MedDiffusion](https://github.com/DiffusionMRI/3D-MedDiffusion).
-
-## Model Details
-
-- **Architecture**: PatchVolume 8× compression with stage 2 training
-- **Compression Factor**: 8× spatial (per dimension)
-- **Latent Channels**: 8
-- **Codebook Size**: 8192 vectors
-- **Input**: CT volumes [B, 1, D, H, W] normalized to [-1, 1]
-- **Output**: Latent embeddings [B, 8, D/8, H/8, W/8]
-
-## Usage
-
-```python
-from models.vqae_wrapper import FrozenVQAE
-
-# Load from HuggingFace Hub
-vae = FrozenVQAE(
-    checkpoint_path="hf://t2ance/ct-vqae-checkpoint/PatchVolume_8x_s2.ckpt",
-    device='cuda'
-)
-
-# Encode CT volume
-ct = torch.randn(1, 1, 200, 128, 128).cuda()
-z = vae.encode(ct)  # [1, 8, 25, 16, 16]
-
-# Decode back
-ct_recon = vae.decode(z)  # [1, 1, 200, 128, 128]
-```
-
 ## Original Source
 
-This checkpoint is from the 3D-MedDiffusion project:
-- Paper: [Diffusion Models for Medical Image Analysis: A Comprehensive Survey](https://arxiv.org/abs/2211.07804)
-- Code: https://github.com/DiffusionMRI/3D-MedDiffusion
-
-## License
+This checkpoint is a direct copy from the 3D-MedDiffusion project:
+- https://github.com/ShanghaiTech-IMPACT/3D-MedDiffusion
 
-Apache 2.0 (same as original 3D-MedDiffusion project)