Update README.md

README.md

@@ -44,22 +44,6 @@ Download all checkpoints:
 huggingface-cli download YilaiLiu-HKU/SlotMem --local-dir . --include "ckpt/*"
 ```
 
-## Installation
-
-```bash
-git clone https://github.com/YilaiLiu-HKU/SlotMem.git
-cd SlotMem
-
-conda create -n slotmem python=3.10 -y
-conda activate slotmem
-
-pip install torch==2.7.1 torchvision==0.22.1 torchaudio==2.7.1 --index-url https://download.pytorch.org/whl/cu128
-pip install -e .
-pip install -r requirements_slotmem.txt
-```
-
-Set `CKPT_DIR` to your local Wan2.2 I2V base model directory.
-
 ## Usage
 
 Download the SlotMem checkpoints into the code repository:
@@ -68,38 +52,8 @@ Download the SlotMem checkpoints into the code repository:
 huggingface-cli download YilaiLiu-HKU/SlotMem --local-dir . --include "ckpt/*"
 ```
 
-Stage-1 inference:
-
-```bash
-CUDA_VISIBLE_DEVICES=0 \
-CKPT_DIR=/path/to/Wan2.2-I2V-A14B \
-JSON_PATH=./sample/test/3_271/rewrite_caption.json \
-REF_IMAGE_PATH=./sample/test/3_271/frame.jpg \
-LOW_EXPERT_CKPT_PATH=./ckpt/stage1/stage1_low.pt \
-HIGH_EXPERT_CKPT_PATH=./ckpt/stage1/stage1_high.pt \
-OUTPUT_ROOT=./inference_outputs/slotmem_stage1 \
-bash test_slotmem_stage1.sh
-```
-
-Stage-2 inference:
-
-```bash
-CUDA_VISIBLE_DEVICES=0 \
-CKPT_DIR=/path/to/Wan2.2-I2V-A14B \
-JSON_PATH=./sample/test/3_271/rewrite_caption.json \
-REF_IMAGE_PATH=./sample/test/3_271/frame.jpg \
-LOW_EXPERT_CKPT_PATH=./ckpt/stage1/stage1_low.pt,./ckpt/stage2/stage2_low.pt \
-HIGH_EXPERT_CKPT_PATH=./ckpt/stage1/stage1_high.pt,./ckpt/stage2/stage2_high.pt \
-OUTPUT_ROOT=./inference_outputs/slotmem_stage2 \
-bash test_slotmem_stage2.sh
-```
-
 For training, data curation, benchmarking, and sample data format, please refer to the [GitHub repository](https://github.com/YilaiLiu-HKU/SlotMem).
 
-## Limitations
-
-SlotMem is designed for research on narrative long video generation and character consistency. The released checkpoints are trained with limited data scaling and may not generalize to every visual domain, character style, or deployment setting. Users should validate outputs carefully and disclose synthetic content in real-world use.
-
 ## Citation
 
 The BibTeX entry will be updated when the paper is public.