Update README.md

README.md

@@ -162,30 +162,20 @@ bash configs_scripts/large/eval/inbatch/run_eval.sh
 
 ## Model Checkpoints
 
-We provide GENIUS model checkpoints in the 🤗 [Hugging Face repository](https://huggingface.co/Sungyeon/GENIUS):
-
-### Stage 1: Residual Quantization Model
-- **Model**: [`rq_clip_large.pth`](https://huggingface.co/Sungyeon/GENIUS/blob/main/checkpoint/rq_clip_large.pth)
-- **Description**: Learns to encode multimodal data into discrete IDs through residual quantization
-- **Size**: ~1.2GB
-
-### Stage 2: Generator Model
-- **Model**: [`GENIUS_t5small.pth`](https://huggingface.co/Sungyeon/GENIUS/blob/main/checkpoint/GENIUS_t5small.pth)
-- **Description**: T5-based sequence-to-sequence model that generates target IDs for retrieval
-- **Size**: ~500MB
-
-### Stage 0: CLIP-SF Model
-- **Model**: [`clip_sf_large.pth`](https://huggingface.co/TIGER-Lab/UniIR/blob/main/checkpoint/CLIP_SF/clip_sf_large.pth)
-- **Source**: [TIGER-Lab/UniIR](https://huggingface.co/TIGER-Lab/UniIR)
-- **Description**: Score-fusion CLIP model for multimodal feature extraction
+We provide model checkpoints for GENIUS in the 🤗 [Hugging Face](https://huggingface.co/Sungyeon/GENIUS):
 
+### How to Download
 ```bash
-# Clone the repository
-git clone https://huggingface.co/Sungyeon/GENIUS
-
-# Download CLIP-SF model
+# Download the CLIP-SF model (Stage 0)
 wget https://huggingface.co/TIGER-Lab/UniIR/resolve/main/checkpoint/CLIP_SF/clip_sf_large.pth -O checkpoint/CLIP_SF/clip_sf_large.pth
+
+# Clone the GENIUS checkpoints (Stage 1 and 2)
+git clone https://huggingface.co/Sungyeon/GENIUS
 ```
+### Each Checkpoints
+- **CLIP-SF Model** (Stage 0): [`clip_sf_large.pth`](https://huggingface.co/TIGER-Lab/UniIR/blob/main/checkpoint/CLIP_SF/clip_sf_large.pth)
+- **Residual Quantization Model** (Stage 1): [`rq_clip_large.pth`](https://huggingface.co/Sungyeon/GENIUS/blob/main/checkpoint/rq_clip_large.pth)
+- **Generator Model** (Stage 2): [`GENIUS_t5small.pth`](https://huggingface.co/Sungyeon/GENIUS/blob/main/checkpoint/GENIUS_t5small.pth)
 
 > Note: All three models are required for full functionality. The CLIP-SF model is used for feature extraction, the Residual Quantization model for ID encoding, and the Generator model for retrieval.