# Clone this repo git clone https://github.com/MINT-SJTU/Evo-1.git cd Evo-1/ # Create a Conda environment conda create -n Evo1 python=3.10 -y conda activate Evo1 # Install requirements cd Evo_1 pip install -r requirements.txt # You may need to reduce MAX_JOBS to suit your computer # (!!! This is a critical step — skipping it may cause lower success rate or unstable robot motion !!!) MAX_JOBS=64 pip install -v flash-attn --no-build-isolation

conda create -n metaworld python=3.10 -y conda activate metaworld pip install mujoco pip install metaworld pip install websockets pip install opencv-python pip install packaging pip install huggingface_hub

hf download MINT-SJTU/Evo1_MetaWorld --local-dir /path/to/save/checkpoint/

# Terminal 1 conda activate Evo1 cd Evo_1 python scripts/Evo1_server.py

# Terminal 2 conda activate metaworld cd MetaWorld_evaluation python mt50_evo1_client_prompt.py

conda create -n libero python=3.8.13 -y conda activate libero cd LIBERO_evaluation/ git clone https://github.com/Lifelong-Robot-Learning/LIBERO.git cd LIBERO pip install -r requirements.txt pip install torch==1.11.0+cu113 torchvision==0.12.0+cu113 torchaudio==0.11.0 --extra-index-url https://download.pytorch.org/whl/cu113 pip install -e . pip install websockets pip install huggingface_hub

hf download MINT-SJTU/Evo1_LIBERO --local-dir /path/to/save/checkpoint/

# Terminal 1 conda activate Evo1 cd Evo_1 python scripts/Evo1_server.py

# Terminal 2 conda activate libero cd LIBERO_evaluation python libero_client_4tasks.py

mkdir Evo1_training_dataset/ cd Evo1_training_dataset/ GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/datasets/MINT-SJTU/Evo1_MetaWorld_Dataset cd Evo1_MetaWorld_Dataset/ git lfs pull

accelerate config 

conda activate Evo1 cd Evo_1/ accelerate launch --num_processes 1 --num_machines 1 --deepspeed_config_file ds_config.json scripts/train.py --run_name Evo1_metaworld_stage1 --action_head flowmatching --use_augmentation --lr 1e-5 --dropout 0.2 --weight_decay 1e-3 --batch_size 16 --image_size 448 --max_steps 5000 --log_interval 10 --ckpt_interval 2500 --warmup_steps 1000 --grad_clip_norm 1.0 --num_layers 8 --horizon 50 --finetune_action_head --disable_wandb --vlm_name OpenGVLab/InternVL3-1B --dataset_config_path dataset/config.yaml --per_action_dim 24 --state_dim 24 --save_dir /your/path/checkpoints/stage1

conda activate Evo1 cd Evo_1/ accelerate launch --num_processes 1 --num_machines 1 --deepspeed_config_file ds_config.json scripts/train.py --run_name Evo1_metaworld_stage2 --action_head flowmatching --use_augmentation --lr 1e-5 --dropout 0.2 --weight_decay 1e-3 --batch_size 16 --image_size 448 --max_steps 80000 --log_interval 10 --ckpt_interval 2500 --warmup_steps 1000 --grad_clip_norm 1.0 --num_layers 8 --horizon 50 --finetune_vlm --finetune_action_head --disable_wandb --vlm_name OpenGVLab/InternVL3-1B --dataset_config_path dataset/config.yaml --per_action_dim 24 --state_dim 24 --save_dir /your/path/checkpoints/stage2 --resume --resume_pretrain --resume_path /your/path/checkpoints/stage1/step_5000

accelerate launch --num_processes 1 --num_machines 1 --deepspeed_config_file ds_config.json scripts/train.py --run_name Your_own_name --action_head flowmatching --use_augmentation --lr 1e-5 --dropout 0.2 --weight_decay 1e-3 --batch_size 16 --image_size 448 --max_steps 80000 --log_interval 10 --ckpt_interval 2500 --warmup_steps 1000 --grad_clip_norm 1.0 --num_layers 8 --horizon 50 --finetune_vlm --finetune_action_head --disable_wandb --vlm_name OpenGVLab/InternVL3-1B --dataset_config_path dataset/config.yaml --per_action_dim 24 --state_dim 24 --save_dir /your/path/to/save/the/checkpoints/ --resume --resume_path /the/checkpoint/path/you/want/to/resume/from/step_20000

 obs = { # You need to change the image size to 448x448 before send in obs "image": [base_proc.tolist(), wrist_proc.tolist(), dummy_proc.tolist()], # This shows which image is valid. "image_mask": [int(i) for i in [1, 1, 0]], # This is the state of the robot. "state": state.astype(float).tolist(), # This is the action mask that shows which action is valid. "action_mask": [[int(i) for i in action_mask[0]]], # This is the instruction of the task "prompt": task_instruction } try: # Send the observation to the server await ws.send(json.dumps(obs)) result = await ws.recv() # Get the action chunk action_chunk = torch.tensor(json.loads(result)) except Exception as e: print(f"❌ Inference Error: {e}") await asyncio.sleep(0.5) continue 

# Create and activate the conda environment for data collection conda create -y -n lerobot python=3.10 conda activate lerobot # Clone the LeRobot repository git clone https://github.com/huggingface/lerobot.git cd lerobot # Checkout the version compatible with v2.1 data format git checkout v0.3.2 pip install -e . pip install -e ".[feetech]"

#Prepare the environment for Evo1_SO100 cd Evo_1/so100_evo1/ conda create -n Evo1_SO100 python=3.10 conda activate Evo1_SO100 #Install FlashAttention wget https://ghproxy.net/https://github.com/Dao-AILab/flash-attention/releases/download/v2.8.3/flash_attn-2.8.3+cu12torch2.7cxx11abiTRUE-cp310-cp310-linux_x86_64.whl pip install flash_attn-2.8.3+cu12torch2.7cxx11abiTRUE-cp310-cp310-linux_x86_64.whl #Install LeRobot conda install ffmpeg -c conda-forge cd lerobot-main pip install -e. pip install -e ".[feetech]" cd Evo_1/so100_evo1/ #Set your own LEROBOT_HOME which include the calibration file of so100 export HF_LEROBOT_HOME="Adress of your own LEROBOT_HOME" pip install transformers accelerate pip install timm

hf download MINT-SJTU/Evo1_SO100 --local-dir /path/to/save/checkpoint/

#Run the command cd Evo-1/so100_evo1 lerobot-record \ --robot.type=so100_follower \ --robot.port=/dev/ttyACMXXXXXXX \ --robot.id=your_so100_follower_arm_id \ --robot.cameras="{   front: {type: opencv, index_or_path: 2, width: 640, height: 480, fps: 30},  wrist: {type: opencv, index_or_path: 0, width: 640, height: 480, fps: 30}  }" \ --display_data=true \ --dataset.repo_id=${HF_USER}/eval_evo1 \ --dataset.single_task="prompt of your task" \ --policy.path= /path/of/your/checkpoint/ #Command example lerobot-record \ --robot.type=so100_follower \ --robot.port=/dev/ttyACM1 \ --robot.id=new_follower_arm \ --robot.cameras="{   front_env: {type: opencv, index_or_path: 2, width: 640, height: 480, fps: 30},  side_env: {type: opencv, index_or_path: 0, width: 640, height: 480, fps: 30}  }" \ --display_data=true \ --dataset.repo_id=yinxinyuchen/eval_evo1 \ --dataset.single_task="Grab the green cube and put the cube in the green box" \ --policy.path=/home/dell/step_20000/

@article{lin2025evo, title={Evo-1: Lightweight Vision-Language-Action Model with Preserved Semantic Alignment}, author={Lin, Tao and Zhong, Yilei and Du, Yuxin and Zhang, Jingjing and Liu, Jiting and Chen, Yinxinyu and Gu, Encheng and Liu, Ziyan and Cai, Hongyi and Zou, Yanwen and others}, journal={arXiv preprint arXiv:2511.04555}, year={2025} }