init

.gitignore

@@ -128,3 +128,4 @@ Data/Pretrained
 Data/utils.py
 Experiment/checkpoint
 Experiment/log
+ckpts/unifolm_wma_dual.ckpt

configs/inference/world_model_interaction.yaml

@@ -222,7 +222,7 @@ data:
     test:
       target: unifolm_wma.data.wma_data.WMAData
       params:
-        data_dir: '/path/to/unifolm-world-model-action/examples/world_model_interaction_prompts'
+        data_dir: '/home/qhy/unifolm-world-model-action/examples/world_model_interaction_prompts'
         video_length: ${model.params.wma_config.params.temporal_length}
         frame_stride: 2
         load_raw_resolution: True

psnr_score_for_challenge.py

@@ -0,0 +1,89 @@
+import os
+import glob
+import numpy as np
+import json
+from argparse import ArgumentParser, ArgumentDefaultsHelpFormatter
+from tqdm import tqdm
+from moviepy.video.io.VideoFileClip import VideoFileClip
+import PIL.Image
+
+
+def calculate_psnr(img1, img2):
+    mse = np.mean((img1.astype(np.float64) - img2.astype(np.float64)) ** 2)
+    if mse == 0:
+        return float('inf')
+    max_pixel = 255.0
+    psnr = 20 * np.log10(max_pixel / np.sqrt(mse))
+    return psnr
+
+
+def process_video_psnr(gt_path, pred_path):
+    try:
+        clip_gt = VideoFileClip(gt_path)
+        clip_pred = VideoFileClip(pred_path)
+
+        fps = min(clip_gt.fps, clip_pred.fps)
+        duration = min(clip_gt.duration, clip_pred.duration)
+
+        time_points = np.arange(0, duration, 1.0 / fps)
+
+        video_psnrs = []
+
+        for t in time_points:
+            frame_gt = clip_gt.get_frame(t)
+            frame_pred = clip_pred.get_frame(t)
+
+            img_gt = PIL.Image.fromarray(frame_gt).resize((256, 256), PIL.Image.Resampling.BILINEAR)
+            img_pred = PIL.Image.fromarray(frame_pred).resize((256, 256), PIL.Image.Resampling.BILINEAR)
+
+            psnr = calculate_psnr(np.array(img_gt), np.array(img_pred))
+            video_psnrs.append(psnr)
+
+        clip_gt.close()
+        clip_pred.close()
+
+        return np.mean(video_psnrs) if video_psnrs else 0.0
+
+    except Exception as e:
+        print(f"Error processing {os.path.basename(gt_path)}: {e}")
+        return None
+
+
+def main():
+    parser = ArgumentParser(formatter_class=ArgumentDefaultsHelpFormatter)
+    parser.add_argument('--gt_video', type=str, required=True, help='path to reference videos')
+    parser.add_argument('--pred_video', type=str, required=True, help='path to pred videos')
+    parser.add_argument('--output_file', type=str, default=None, help='path to output file')
+    args = parser.parse_args()
+
+    if not os.path.exists(args.gt_video):
+        print(f"Error: GT video not found at {args.gt_video}")
+        return
+    if not os.path.exists(args.pred_video):
+        print(f"Error: Pred video not found at {args.pred_video}")
+        return
+
+    print(f"Comparing:\nRef: {args.gt_video}\nPred: {args.pred_video}")
+
+    v_psnr = process_video_psnr(args.gt_video, args.pred_video)
+
+    if v_psnr is not None:
+        print("-" * 30)
+        print(f"Video PSNR: {v_psnr:.4f} dB")
+        print("-" * 30)
+
+        if args.output_file:
+            result = {
+                "gt_video": args.gt_video,
+                "pred_video": args.pred_video,
+                "psnr": v_psnr
+            }
+            with open(args.output_file, 'w') as f:
+                json.dump(result, f, indent=4)
+            print(f"Result saved to {args.output_file}")
+    else:
+        print("Failed to calculate PSNR.")
+
+
+if __name__ == '__main__':
+    main()

unitree_g1_pack_camera/case1/run_world_model_interaction.sh

@@ -0,0 +1,24 @@
+res_dir="unitree_g1_pack_camera/case1"
+dataset="unitree_g1_pack_camera"
+
+{
+    time CUDA_VISIBLE_DEVICES=0 python3 scripts/evaluation/world_model_interaction.py \
+        --seed 123 \
+        --ckpt_path ckpts/unifolm_wma_dual.ckpt \
+        --config configs/inference/world_model_interaction.yaml \
+        --savedir "${res_dir}/output" \
+        --bs 1 --height 320 --width 512 \
+        --unconditional_guidance_scale 1.0 \
+        --ddim_steps 50 \
+        --ddim_eta 1.0 \
+        --prompt_dir "unitree_g1_pack_camera/case1/world_model_interaction_prompts" \
+        --dataset ${dataset} \
+        --video_length 16 \
+        --frame_stride 6 \
+        --n_action_steps 16 \
+        --exe_steps 16 \
+        --n_iter 11 \
+        --timestep_spacing 'uniform_trailing' \
+        --guidance_rescale 0.7 \
+        --perframe_ae
+} 2>&1 | tee "${res_dir}/output.log"

unitree_g1_pack_camera/case1/world_model_interaction_prompts/unitree_g1_pack_camera.csv

@@ -0,0 +1,2 @@
+videoid,contentUrl,duration,data_dir,instruction,dynamic_confidence,dynamic_wording,dynamic_source_category,embodiment,fps
+0,x,x,unitree_g1_pack_camera,mount camera,x,x,x,G1_Dex1,30

unitree_g1_pack_camera/case2/run_world_model_interaction.sh

@@ -0,0 +1,24 @@
+res_dir="unitree_g1_pack_camera/case2"
+dataset="unitree_g1_pack_camera"
+
+{
+    time CUDA_VISIBLE_DEVICES=0 python3 scripts/evaluation/world_model_interaction.py \
+        --seed 123 \
+        --ckpt_path ckpts/unifolm_wma_dual.ckpt \
+        --config configs/inference/world_model_interaction.yaml \
+        --savedir "${res_dir}/output" \
+        --bs 1 --height 320 --width 512 \
+        --unconditional_guidance_scale 1.0 \
+        --ddim_steps 50 \
+        --ddim_eta 1.0 \
+        --prompt_dir "unitree_g1_pack_camera/case2/world_model_interaction_prompts" \
+        --dataset ${dataset} \
+        --video_length 16 \
+        --frame_stride 6 \
+        --n_action_steps 16 \
+        --exe_steps 16 \
+        --n_iter 11 \
+        --timestep_spacing 'uniform_trailing' \
+        --guidance_rescale 0.7 \
+        --perframe_ae
+} 2>&1 | tee "${res_dir}/output.log"

unitree_g1_pack_camera/case2/world_model_interaction_prompts/unitree_g1_pack_camera.csv

@@ -0,0 +1,2 @@
+videoid,contentUrl,duration,data_dir,instruction,dynamic_confidence,dynamic_wording,dynamic_source_category,embodiment,fps
+50,x,x,unitree_g1_pack_camera,mount camera,x,x,x,G1_Dex1,30

unitree_g1_pack_camera/case3/run_world_model_interaction.sh

@@ -0,0 +1,24 @@
+res_dir="unitree_g1_pack_camera/case3"
+dataset="unitree_g1_pack_camera"
+
+{
+    time CUDA_VISIBLE_DEVICES=0 python3 scripts/evaluation/world_model_interaction.py \
+        --seed 123 \
+        --ckpt_path ckpts/unifolm_wma_dual.ckpt \
+        --config configs/inference/world_model_interaction.yaml \
+        --savedir "${res_dir}/output" \
+        --bs 1 --height 320 --width 512 \
+        --unconditional_guidance_scale 1.0 \
+        --ddim_steps 50 \
+        --ddim_eta 1.0 \
+        --prompt_dir "unitree_g1_pack_camera/case3/world_model_interaction_prompts" \
+        --dataset ${dataset} \
+        --video_length 16 \
+        --frame_stride 6 \
+        --n_action_steps 16 \
+        --exe_steps 16 \
+        --n_iter 11 \
+        --timestep_spacing 'uniform_trailing' \
+        --guidance_rescale 0.7 \
+        --perframe_ae
+} 2>&1 | tee "${res_dir}/output.log"

unitree_g1_pack_camera/case3/world_model_interaction_prompts/unitree_g1_pack_camera.csv

@@ -0,0 +1,2 @@
+videoid,contentUrl,duration,data_dir,instruction,dynamic_confidence,dynamic_wording,dynamic_source_category,embodiment,fps
+100,x,x,unitree_g1_pack_camera,mount camera,x,x,x,G1_Dex1,30

unitree_g1_pack_camera/case4/run_world_model_interaction.sh

@@ -0,0 +1,24 @@
+res_dir="unitree_g1_pack_camera/case4"
+dataset="unitree_g1_pack_camera"
+
+{
+    time CUDA_VISIBLE_DEVICES=0 python3 scripts/evaluation/world_model_interaction.py \
+        --seed 123 \
+        --ckpt_path ckpts/unifolm_wma_dual.ckpt \
+        --config configs/inference/world_model_interaction.yaml \
+        --savedir "${res_dir}/output" \
+        --bs 1 --height 320 --width 512 \
+        --unconditional_guidance_scale 1.0 \
+        --ddim_steps 50 \
+        --ddim_eta 1.0 \
+        --prompt_dir "unitree_g1_pack_camera/case4/world_model_interaction_prompts" \
+        --dataset ${dataset} \
+        --video_length 16 \
+        --frame_stride 6 \
+        --n_action_steps 16 \
+        --exe_steps 16 \
+        --n_iter 11 \
+        --timestep_spacing 'uniform_trailing' \
+        --guidance_rescale 0.7 \
+        --perframe_ae
+} 2>&1 | tee "${res_dir}/output.log"

unitree_g1_pack_camera/case4/world_model_interaction_prompts/unitree_g1_pack_camera.csv

@@ -0,0 +1,2 @@
+videoid,contentUrl,duration,data_dir,instruction,dynamic_confidence,dynamic_wording,dynamic_source_category,embodiment,fps
+200,x,x,unitree_g1_pack_camera,mount camera,x,x,x,G1_Dex1,30

unitree_z1_dual_arm_cleanup_pencils/case1/run_world_model_interaction.sh

@@ -0,0 +1,24 @@
+res_dir="unitree_z1_dual_arm_cleanup_pencils/case1"
+dataset="unitree_z1_dual_arm_cleanup_pencils"
+
+{
+    time CUDA_VISIBLE_DEVICES=0 python3 scripts/evaluation/world_model_interaction.py \
+        --seed 123 \
+        --ckpt_path ckpts/unifolm_wma_dual.ckpt \
+        --config configs/inference/world_model_interaction.yaml \
+        --savedir "${res_dir}/output" \
+        --bs 1 --height 320 --width 512 \
+        --unconditional_guidance_scale 1.0 \
+        --ddim_steps 50 \
+        --ddim_eta 1.0 \
+        --prompt_dir "unitree_z1_dual_arm_cleanup_pencils/case1/world_model_interaction_prompts" \
+        --dataset ${dataset} \
+        --video_length 16 \
+        --frame_stride 4 \
+        --n_action_steps 16 \
+        --exe_steps 16 \
+        --n_iter 8 \
+        --timestep_spacing 'uniform_trailing' \
+        --guidance_rescale 0.7 \
+        --perframe_ae
+} 2>&1 | tee "${res_dir}/output.log"

unitree_z1_dual_arm_cleanup_pencils/case1/world_model_interaction_prompts/unitree_z1_dual_arm_cleanup_pencils.csv

@@ -0,0 +1,2 @@
+videoid,contentUrl,duration,data_dir,instruction,dynamic_confidence,dynamic_wording,dynamic_source_category,embodiment,fps
+0,x,x,unitree_z1_dual_arm_cleanup_pencils,clean up eraser and pencils,x,x,x,Z1_Dual_Dex1,30

unitree_z1_dual_arm_cleanup_pencils/case2/run_world_model_interaction.sh

@@ -0,0 +1,24 @@
+res_dir="unitree_z1_dual_arm_cleanup_pencils/case2"
+dataset="unitree_z1_dual_arm_cleanup_pencils"
+
+{
+    time CUDA_VISIBLE_DEVICES=0 python3 scripts/evaluation/world_model_interaction.py \
+        --seed 123 \
+        --ckpt_path ckpts/unifolm_wma_dual.ckpt \
+        --config configs/inference/world_model_interaction.yaml \
+        --savedir "${res_dir}/output" \
+        --bs 1 --height 320 --width 512 \
+        --unconditional_guidance_scale 1.0 \
+        --ddim_steps 50 \
+        --ddim_eta 1.0 \
+        --prompt_dir "unitree_z1_dual_arm_cleanup_pencils/case2/world_model_interaction_prompts" \
+        --dataset ${dataset} \
+        --video_length 16 \
+        --frame_stride 4 \
+        --n_action_steps 16 \
+        --exe_steps 16 \
+        --n_iter 8 \
+        --timestep_spacing 'uniform_trailing' \
+        --guidance_rescale 0.7 \
+        --perframe_ae
+} 2>&1 | tee "${res_dir}/output.log"

unitree_z1_dual_arm_cleanup_pencils/case2/world_model_interaction_prompts/unitree_z1_dual_arm_cleanup_pencils.csv

@@ -0,0 +1,2 @@
+videoid,contentUrl,duration,data_dir,instruction,dynamic_confidence,dynamic_wording,dynamic_source_category,embodiment,fps
+50,x,x,unitree_z1_dual_arm_cleanup_pencils,clean up eraser and pencils,x,x,x,Z1_Dual_Dex1,30

unitree_z1_dual_arm_cleanup_pencils/case3/run_world_model_interaction.sh

@@ -0,0 +1,24 @@
+res_dir="unitree_z1_dual_arm_cleanup_pencils/case3"
+dataset="unitree_z1_dual_arm_cleanup_pencils"
+
+{
+    time CUDA_VISIBLE_DEVICES=0 python3 scripts/evaluation/world_model_interaction.py \
+        --seed 123 \
+        --ckpt_path ckpts/unifolm_wma_dual.ckpt \
+        --config configs/inference/world_model_interaction.yaml \
+        --savedir "${res_dir}/output" \
+        --bs 1 --height 320 --width 512 \
+        --unconditional_guidance_scale 1.0 \
+        --ddim_steps 50 \
+        --ddim_eta 1.0 \
+        --prompt_dir "unitree_z1_dual_arm_cleanup_pencils/case3/world_model_interaction_prompts" \
+        --dataset ${dataset} \
+        --video_length 16 \
+        --frame_stride 4 \
+        --n_action_steps 16 \
+        --exe_steps 16 \
+        --n_iter 8 \
+        --timestep_spacing 'uniform_trailing' \
+        --guidance_rescale 0.7 \
+        --perframe_ae
+} 2>&1 | tee "${res_dir}/output.log"

unitree_z1_dual_arm_cleanup_pencils/case3/world_model_interaction_prompts/unitree_z1_dual_arm_cleanup_pencils.csv

@@ -0,0 +1,2 @@
+videoid,contentUrl,duration,data_dir,instruction,dynamic_confidence,dynamic_wording,dynamic_source_category,embodiment,fps
+100,x,x,unitree_z1_dual_arm_cleanup_pencils,clean up eraser and pencils,x,x,x,Z1_Dual_Dex1,30

unitree_z1_dual_arm_cleanup_pencils/case4/run_world_model_interaction.sh

@@ -0,0 +1,24 @@
+res_dir="unitree_z1_dual_arm_cleanup_pencils/case4"
+dataset="unitree_z1_dual_arm_cleanup_pencils"
+
+{
+    time CUDA_VISIBLE_DEVICES=0 python3 scripts/evaluation/world_model_interaction.py \
+        --seed 123 \
+        --ckpt_path ckpts/unifolm_wma_dual.ckpt \
+        --config configs/inference/world_model_interaction.yaml \
+        --savedir "${res_dir}/output" \
+        --bs 1 --height 320 --width 512 \
+        --unconditional_guidance_scale 1.0 \
+        --ddim_steps 50 \
+        --ddim_eta 1.0 \
+        --prompt_dir "unitree_z1_dual_arm_cleanup_pencils/case4/world_model_interaction_prompts" \
+        --dataset ${dataset} \
+        --video_length 16 \
+        --frame_stride 4 \
+        --n_action_steps 16 \
+        --exe_steps 16 \
+        --n_iter 8 \
+        --timestep_spacing 'uniform_trailing' \
+        --guidance_rescale 0.7 \
+        --perframe_ae
+} 2>&1 | tee "${res_dir}/output.log"

unitree_z1_dual_arm_cleanup_pencils/case4/world_model_interaction_prompts/unitree_z1_dual_arm_cleanup_pencils.csv

@@ -0,0 +1,2 @@
+videoid,contentUrl,duration,data_dir,instruction,dynamic_confidence,dynamic_wording,dynamic_source_category,embodiment,fps
+200,x,x,unitree_z1_dual_arm_cleanup_pencils,clean up eraser and pencils,x,x,x,Z1_Dual_Dex1,30

unitree_z1_dual_arm_stackbox/case1/run_world_model_interaction.sh

@@ -0,0 +1,24 @@
+res_dir="unitree_z1_dual_arm_stackbox/case1"
+dataset="unitree_z1_dual_arm_stackbox"
+
+{
+    time CUDA_VISIBLE_DEVICES=0 python3 scripts/evaluation/world_model_interaction.py \
+        --seed 123 \
+        --ckpt_path ckpts/unifolm_wma_dual.ckpt \
+        --config configs/inference/world_model_interaction.yaml \
+        --savedir "${res_dir}/output" \
+        --bs 1 --height 320 --width 512 \
+        --unconditional_guidance_scale 1.0 \
+        --ddim_steps 50 \
+        --ddim_eta 1.0 \
+        --prompt_dir "unitree_z1_dual_arm_stackbox/case1/world_model_interaction_prompts" \
+        --dataset ${dataset} \
+        --video_length 16 \
+        --frame_stride 4 \
+        --n_action_steps 16 \
+        --exe_steps 16 \
+        --n_iter 7 \
+        --timestep_spacing 'uniform_trailing' \
+        --guidance_rescale 0.7 \
+        --perframe_ae
+} 2>&1 | tee "${res_dir}/output.log"

unitree_z1_dual_arm_stackbox/case1/world_model_interaction_prompts/unitree_z1_dual_arm_stackbox.csv

@@ -0,0 +1,2 @@
+videoid,contentUrl,duration,data_dir,instruction,dynamic_confidence,dynamic_wording,dynamic_source_category,embodiment,fps
+5,x,x,unitree_z1_dual_arm_stackbox,"Pick up the red cup on the table.",x,x,x,Unitree Z1 Robot Dual-Arm,30

unitree_z1_dual_arm_stackbox/case2/run_world_model_interaction.sh

@@ -0,0 +1,24 @@
+res_dir="unitree_z1_dual_arm_stackbox/case2"
+dataset="unitree_z1_dual_arm_stackbox"
+
+{
+    time CUDA_VISIBLE_DEVICES=0 python3 scripts/evaluation/world_model_interaction.py \
+        --seed 123 \
+        --ckpt_path ckpts/unifolm_wma_dual.ckpt \
+        --config configs/inference/world_model_interaction.yaml \
+        --savedir "${res_dir}/output" \
+        --bs 1 --height 320 --width 512 \
+        --unconditional_guidance_scale 1.0 \
+        --ddim_steps 50 \
+        --ddim_eta 1.0 \
+        --prompt_dir "unitree_z1_dual_arm_stackbox/case2/world_model_interaction_prompts" \
+        --dataset ${dataset} \
+        --video_length 16 \
+        --frame_stride 4 \
+        --n_action_steps 16 \
+        --exe_steps 16 \
+        --n_iter 7 \
+        --timestep_spacing 'uniform_trailing' \
+        --guidance_rescale 0.7 \
+        --perframe_ae
+} 2>&1 | tee "${res_dir}/output.log"

unitree_z1_dual_arm_stackbox/case2/world_model_interaction_prompts/unitree_z1_dual_arm_stackbox.csv

@@ -0,0 +1,2 @@
+videoid,contentUrl,duration,data_dir,instruction,dynamic_confidence,dynamic_wording,dynamic_source_category,embodiment,fps
+15,x,x,unitree_z1_dual_arm_stackbox,"Pick up the red cup on the table.",x,x,x,Unitree Z1 Robot Dual-Arm,30

unitree_z1_dual_arm_stackbox/case3/run_world_model_interaction.sh

@@ -0,0 +1,24 @@
+res_dir="unitree_z1_dual_arm_stackbox/case3"
+dataset="unitree_z1_dual_arm_stackbox"
+
+{
+    time CUDA_VISIBLE_DEVICES=0 python3 scripts/evaluation/world_model_interaction.py \
+        --seed 123 \
+        --ckpt_path ckpts/unifolm_wma_dual.ckpt \
+        --config configs/inference/world_model_interaction.yaml \
+        --savedir "${res_dir}/output" \
+        --bs 1 --height 320 --width 512 \
+        --unconditional_guidance_scale 1.0 \
+        --ddim_steps 50 \
+        --ddim_eta 1.0 \
+        --prompt_dir "unitree_z1_dual_arm_stackbox/case3/world_model_interaction_prompts" \
+        --dataset ${dataset} \
+        --video_length 16 \
+        --frame_stride 4 \
+        --n_action_steps 16 \
+        --exe_steps 16 \
+        --n_iter 7 \
+        --timestep_spacing 'uniform_trailing' \
+        --guidance_rescale 0.7 \
+        --perframe_ae
+} 2>&1 | tee "${res_dir}/output.log"

unitree_z1_dual_arm_stackbox/case3/world_model_interaction_prompts/unitree_z1_dual_arm_stackbox.csv

@@ -0,0 +1,2 @@
+videoid,contentUrl,duration,data_dir,instruction,dynamic_confidence,dynamic_wording,dynamic_source_category,embodiment,fps
+25,x,x,unitree_z1_dual_arm_stackbox,"Pick up the red cup on the table.",x,x,x,Unitree Z1 Robot Dual-Arm,30

unitree_z1_dual_arm_stackbox/case4/run_world_model_interaction.sh

@@ -0,0 +1,24 @@
+res_dir="unitree_z1_dual_arm_stackbox/case4"
+dataset="unitree_z1_dual_arm_stackbox"
+
+{
+    time CUDA_VISIBLE_DEVICES=0 python3 scripts/evaluation/world_model_interaction.py \
+        --seed 123 \
+        --ckpt_path ckpts/unifolm_wma_dual.ckpt \
+        --config configs/inference/world_model_interaction.yaml \
+        --savedir "${res_dir}/output" \
+        --bs 1 --height 320 --width 512 \
+        --unconditional_guidance_scale 1.0 \
+        --ddim_steps 50 \
+        --ddim_eta 1.0 \
+        --prompt_dir "unitree_z1_dual_arm_stackbox/case4/world_model_interaction_prompts" \
+        --dataset ${dataset} \
+        --video_length 16 \
+        --frame_stride 4 \
+        --n_action_steps 16 \
+        --exe_steps 16 \
+        --n_iter 7 \
+        --timestep_spacing 'uniform_trailing' \
+        --guidance_rescale 0.7 \
+        --perframe_ae
+} 2>&1 | tee "${res_dir}/output.log"

unitree_z1_dual_arm_stackbox/case4/world_model_interaction_prompts/unitree_z1_dual_arm_stackbox.csv

@@ -0,0 +1,2 @@
+videoid,contentUrl,duration,data_dir,instruction,dynamic_confidence,dynamic_wording,dynamic_source_category,embodiment,fps
+35,x,x,unitree_z1_dual_arm_stackbox,"Pick up the red cup on the table.",x,x,x,Unitree Z1 Robot Dual-Arm,30

unitree_z1_dual_arm_stackbox_v2/case1/run_world_model_interaction.sh

@@ -0,0 +1,24 @@
+res_dir="unitree_z1_dual_arm_stackbox_v2/case1"
+dataset="unitree_z1_dual_arm_stackbox_v2"
+
+{
+    time CUDA_VISIBLE_DEVICES=0 python3 scripts/evaluation/world_model_interaction.py \
+        --seed 123 \
+        --ckpt_path ckpts/unifolm_wma_dual.ckpt \
+        --config configs/inference/world_model_interaction.yaml \
+        --savedir "${res_dir}/output" \
+        --bs 1 --height 320 --width 512 \
+        --unconditional_guidance_scale 1.0 \
+        --ddim_steps 50 \
+        --ddim_eta 1.0 \
+        --prompt_dir "unitree_z1_dual_arm_stackbox_v2/case1/world_model_interaction_prompts" \
+        --dataset ${dataset} \
+        --video_length 16 \
+        --frame_stride 4 \
+        --n_action_steps 16 \
+        --exe_steps 16 \
+        --n_iter 11 \
+        --timestep_spacing 'uniform_trailing' \
+        --guidance_rescale 0.7 \
+        --perframe_ae
+} 2>&1 | tee "${res_dir}/output.log"

unitree_z1_dual_arm_stackbox_v2/case1/world_model_interaction_prompts/unitree_z1_dual_arm_stackbox_v2.csv

@@ -0,0 +1,2 @@
+videoid,contentUrl,duration,data_dir,instruction,dynamic_confidence,dynamic_wording,dynamic_source_category,embodiment,fps
+5,x,x,unitree_z1_dual_arm_stackbox_v2,"Stack the blocks in the rectangular block: red at the bottom, yellow in the middle, green on top",x,x,x,Unitree Z1 Robot Dual-Arm,30

unitree_z1_dual_arm_stackbox_v2/case2/run_world_model_interaction.sh

@@ -0,0 +1,24 @@
+res_dir="unitree_z1_dual_arm_stackbox_v2/case2"
+dataset="unitree_z1_dual_arm_stackbox_v2"
+
+{
+    time CUDA_VISIBLE_DEVICES=0 python3 scripts/evaluation/world_model_interaction.py \
+        --seed 123 \
+        --ckpt_path ckpts/unifolm_wma_dual.ckpt \
+        --config configs/inference/world_model_interaction.yaml \
+        --savedir "${res_dir}/output" \
+        --bs 1 --height 320 --width 512 \
+        --unconditional_guidance_scale 1.0 \
+        --ddim_steps 50 \
+        --ddim_eta 1.0 \
+        --prompt_dir "unitree_z1_dual_arm_stackbox_v2/case2/world_model_interaction_prompts" \
+        --dataset ${dataset} \
+        --video_length 16 \
+        --frame_stride 4 \
+        --n_action_steps 16 \
+        --exe_steps 16 \
+        --n_iter 11 \
+        --timestep_spacing 'uniform_trailing' \
+        --guidance_rescale 0.7 \
+        --perframe_ae
+} 2>&1 | tee "${res_dir}/output.log"

unitree_z1_dual_arm_stackbox_v2/case2/world_model_interaction_prompts/unitree_z1_dual_arm_stackbox_v2.csv

@@ -0,0 +1,2 @@
+videoid,contentUrl,duration,data_dir,instruction,dynamic_confidence,dynamic_wording,dynamic_source_category,embodiment,fps
+15,x,x,unitree_z1_dual_arm_stackbox_v2,"Stack the blocks in the rectangular block: red at the bottom, yellow in the middle, green on top",x,x,x,Unitree Z1 Robot Dual-Arm,30

unitree_z1_dual_arm_stackbox_v2/case3/run_world_model_interaction.sh

@@ -0,0 +1,24 @@
+res_dir="unitree_z1_dual_arm_stackbox_v2/case3"
+dataset="unitree_z1_dual_arm_stackbox_v2"
+
+{
+    time CUDA_VISIBLE_DEVICES=0 python3 scripts/evaluation/world_model_interaction.py \
+        --seed 123 \
+        --ckpt_path ckpts/unifolm_wma_dual.ckpt \
+        --config configs/inference/world_model_interaction.yaml \
+        --savedir "${res_dir}/output" \
+        --bs 1 --height 320 --width 512 \
+        --unconditional_guidance_scale 1.0 \
+        --ddim_steps 50 \
+        --ddim_eta 1.0 \
+        --prompt_dir "unitree_z1_dual_arm_stackbox_v2/case3/world_model_interaction_prompts" \
+        --dataset ${dataset} \
+        --video_length 16 \
+        --frame_stride 4 \
+        --n_action_steps 16 \
+        --exe_steps 16 \
+        --n_iter 11 \
+        --timestep_spacing 'uniform_trailing' \
+        --guidance_rescale 0.7 \
+        --perframe_ae
+} 2>&1 | tee "${res_dir}/output.log"

unitree_z1_dual_arm_stackbox_v2/case3/world_model_interaction_prompts/unitree_z1_dual_arm_stackbox_v2.csv

@@ -0,0 +1,2 @@
+videoid,contentUrl,duration,data_dir,instruction,dynamic_confidence,dynamic_wording,dynamic_source_category,embodiment,fps
+25,x,x,unitree_z1_dual_arm_stackbox_v2,"Stack the blocks in the rectangular block: red at the bottom, yellow in the middle, green on top",x,x,x,Unitree Z1 Robot Dual-Arm,30

unitree_z1_dual_arm_stackbox_v2/case4/run_world_model_interaction.sh

@@ -0,0 +1,24 @@
+res_dir="unitree_z1_dual_arm_stackbox_v2/case4"
+dataset="unitree_z1_dual_arm_stackbox_v2"
+
+{
+    time CUDA_VISIBLE_DEVICES=0 python3 scripts/evaluation/world_model_interaction.py \
+        --seed 123 \
+        --ckpt_path ckpts/unifolm_wma_dual.ckpt \
+        --config configs/inference/world_model_interaction.yaml \
+        --savedir "${res_dir}/output" \
+        --bs 1 --height 320 --width 512 \
+        --unconditional_guidance_scale 1.0 \
+        --ddim_steps 50 \
+        --ddim_eta 1.0 \
+        --prompt_dir "unitree_z1_dual_arm_stackbox_v2/case4/world_model_interaction_prompts" \
+        --dataset ${dataset} \
+        --video_length 16 \
+        --frame_stride 4 \
+        --n_action_steps 16 \
+        --exe_steps 16 \
+        --n_iter 11 \
+        --timestep_spacing 'uniform_trailing' \
+        --guidance_rescale 0.7 \
+        --perframe_ae
+} 2>&1 | tee "${res_dir}/output.log"

unitree_z1_dual_arm_stackbox_v2/case4/world_model_interaction_prompts/unitree_z1_dual_arm_stackbox_v2.csv

@@ -0,0 +1,2 @@
+videoid,contentUrl,duration,data_dir,instruction,dynamic_confidence,dynamic_wording,dynamic_source_category,embodiment,fps
+35,x,x,unitree_z1_dual_arm_stackbox_v2,"Stack the blocks in the rectangular block: red at the bottom, yellow in the middle, green on top",x,x,x,Unitree Z1 Robot Dual-Arm,30

unitree_z1_stackbox/case1/run_world_model_interaction.sh

@@ -0,0 +1,24 @@
+res_dir="unitree_z1_stackbox/case1"
+dataset="unitree_z1_stackbox"
+
+{
+    time CUDA_VISIBLE_DEVICES=0 python3 scripts/evaluation/world_model_interaction.py \
+        --seed 123 \
+        --ckpt_path ckpts/unifolm_wma_dual.ckpt \
+        --config configs/inference/world_model_interaction.yaml \
+        --savedir "${res_dir}/output" \
+        --bs 1 --height 320 --width 512 \
+        --unconditional_guidance_scale 1.0 \
+        --ddim_steps 50 \
+        --ddim_eta 1.0 \
+        --prompt_dir "unitree_z1_stackbox/case1/world_model_interaction_prompts" \
+        --dataset ${dataset} \
+        --video_length 16 \
+        --frame_stride 4 \
+        --n_action_steps 16 \
+        --exe_steps 16 \
+        --n_iter 12 \
+        --timestep_spacing 'uniform_trailing' \
+        --guidance_rescale 0.7 \
+        --perframe_ae
+} 2>&1 | tee "${res_dir}/output.log"

unitree_z1_stackbox/case1/world_model_interaction_prompts/unitree_z1_stackbox.csv

@@ -0,0 +1,2 @@
+videoid,contentUrl,duration,data_dir,instruction,dynamic_confidence,dynamic_wording,dynamic_source_category,embodiment,fps
+5,x,x,unitree_z1_stackbox,"Pick up the red cup on the table.",x,x,x,Unitree Z1 Robot Arm,30

unitree_z1_stackbox/case2/run_world_model_interaction.sh

@@ -0,0 +1,24 @@
+res_dir="unitree_z1_stackbox/case2"
+dataset="unitree_z1_stackbox"
+
+{
+    time CUDA_VISIBLE_DEVICES=0 python3 scripts/evaluation/world_model_interaction.py \
+        --seed 123 \
+        --ckpt_path ckpts/unifolm_wma_dual.ckpt \
+        --config configs/inference/world_model_interaction.yaml \
+        --savedir "${res_dir}/output" \
+        --bs 1 --height 320 --width 512 \
+        --unconditional_guidance_scale 1.0 \
+        --ddim_steps 50 \
+        --ddim_eta 1.0 \
+        --prompt_dir "unitree_z1_stackbox/case2/world_model_interaction_prompts" \
+        --dataset ${dataset} \
+        --video_length 16 \
+        --frame_stride 4 \
+        --n_action_steps 16 \
+        --exe_steps 16 \
+        --n_iter 12 \
+        --timestep_spacing 'uniform_trailing' \
+        --guidance_rescale 0.7 \
+        --perframe_ae
+} 2>&1 | tee "${res_dir}/output.log"

unitree_z1_stackbox/case2/world_model_interaction_prompts/unitree_z1_stackbox.csv

@@ -0,0 +1,2 @@
+videoid,contentUrl,duration,data_dir,instruction,dynamic_confidence,dynamic_wording,dynamic_source_category,embodiment,fps
+15,x,x,unitree_z1_stackbox,"Pick up the red cup on the table.",x,x,x,Unitree Z1 Robot Arm,30

unitree_z1_stackbox/case3/run_world_model_interaction.sh

@@ -0,0 +1,24 @@
+res_dir="unitree_z1_stackbox/case3"
+dataset="unitree_z1_stackbox"
+
+{
+    time CUDA_VISIBLE_DEVICES=0 python3 scripts/evaluation/world_model_interaction.py \
+        --seed 123 \
+        --ckpt_path ckpts/unifolm_wma_dual.ckpt \
+        --config configs/inference/world_model_interaction.yaml \
+        --savedir "${res_dir}/output" \
+        --bs 1 --height 320 --width 512 \
+        --unconditional_guidance_scale 1.0 \
+        --ddim_steps 50 \
+        --ddim_eta 1.0 \
+        --prompt_dir "unitree_z1_stackbox/case3/world_model_interaction_prompts" \
+        --dataset ${dataset} \
+        --video_length 16 \
+        --frame_stride 4 \
+        --n_action_steps 16 \
+        --exe_steps 16 \
+        --n_iter 12 \
+        --timestep_spacing 'uniform_trailing' \
+        --guidance_rescale 0.7 \
+        --perframe_ae
+} 2>&1 | tee "${res_dir}/output.log"

unitree_z1_stackbox/case3/world_model_interaction_prompts/unitree_z1_stackbox.csv

@@ -0,0 +1,2 @@
+videoid,contentUrl,duration,data_dir,instruction,dynamic_confidence,dynamic_wording,dynamic_source_category,embodiment,fps
+25,x,x,unitree_z1_stackbox,"Pick up the red cup on the table.",x,x,x,Unitree Z1 Robot Arm,30

unitree_z1_stackbox/case4/run_world_model_interaction.sh

@@ -0,0 +1,24 @@
+res_dir="unitree_z1_stackbox/case4"
+dataset="unitree_z1_stackbox"
+
+{
+    time CUDA_VISIBLE_DEVICES=0 python3 scripts/evaluation/world_model_interaction.py \
+        --seed 123 \
+        --ckpt_path ckpts/unifolm_wma_dual.ckpt \
+        --config configs/inference/world_model_interaction.yaml \
+        --savedir "${res_dir}/output" \
+        --bs 1 --height 320 --width 512 \
+        --unconditional_guidance_scale 1.0 \
+        --ddim_steps 50 \
+        --ddim_eta 1.0 \
+        --prompt_dir "unitree_z1_stackbox/case4/world_model_interaction_prompts" \
+        --dataset ${dataset} \
+        --video_length 16 \
+        --frame_stride 4 \
+        --n_action_steps 16 \
+        --exe_steps 16 \
+        --n_iter 12 \
+        --timestep_spacing 'uniform_trailing' \
+        --guidance_rescale 0.7 \
+        --perframe_ae
+} 2>&1 | tee "${res_dir}/output.log"