Files
hao-ai-lab-FastVideo/examples/training/finetune/MatrixGame2.0/validation_df.json
T
2026-03-07 23:22:50 +00:00

325 lines
16 KiB
JSON

{
"data": [
{
"caption": "00",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000051_Bravo_instance_002.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000051_Bravo_instance_002.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "01",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000051_Bravo_instance_003.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000051_Bravo_instance_003.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "02",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000052_Alpha_instance_000.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000052_Alpha_instance_000.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "03",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000052_Alpha_instance_001.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000052_Alpha_instance_001.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "04",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000052_Alpha_instance_003.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000052_Alpha_instance_003.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "05",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000052_Bravo_instance_000.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000052_Bravo_instance_000.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "06",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000052_Bravo_instance_001.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000052_Bravo_instance_001.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "07",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000052_Bravo_instance_003.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000052_Bravo_instance_003.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "08",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000053_Alpha_instance_000.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000053_Alpha_instance_000.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "09",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000053_Alpha_instance_001.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000053_Alpha_instance_001.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "10",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000053_Alpha_instance_002.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000053_Alpha_instance_002.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "11",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000053_Alpha_instance_003.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000053_Alpha_instance_003.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "12",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000053_Bravo_instance_000.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000053_Bravo_instance_000.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "13",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000053_Bravo_instance_001.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000053_Bravo_instance_001.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "14",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000053_Bravo_instance_002.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000053_Bravo_instance_002.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "15",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000053_Bravo_instance_003.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000053_Bravo_instance_003.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "16",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000054_Alpha_instance_000.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000054_Alpha_instance_000.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "17",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000054_Alpha_instance_001.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000054_Alpha_instance_001.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "18",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000054_Alpha_instance_002.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000054_Alpha_instance_002.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "19",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000054_Alpha_instance_003.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000054_Alpha_instance_003.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "20",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000054_Bravo_instance_000.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000054_Bravo_instance_000.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "21",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000054_Bravo_instance_001.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000054_Bravo_instance_001.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "22",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000054_Bravo_instance_002.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000054_Bravo_instance_002.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "23",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000054_Bravo_instance_003.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000054_Bravo_instance_003.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "24",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000055_Alpha_instance_000.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000055_Alpha_instance_000.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "25",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000055_Alpha_instance_001.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000055_Alpha_instance_001.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "26",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000055_Alpha_instance_002.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000055_Alpha_instance_002.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "27",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000055_Alpha_instance_003.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000055_Alpha_instance_003.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "28",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000055_Bravo_instance_000.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000055_Bravo_instance_000.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "29",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000055_Bravo_instance_001.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000055_Bravo_instance_001.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "30",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000055_Bravo_instance_002.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000055_Bravo_instance_002.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "31",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000055_Bravo_instance_003.png",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train/000010/train/75_batch2_split_0_000055_Bravo_instance_003.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
}
]
}