Files
hao-ai-lab-FastVideo/examples/training/finetune/MatrixGame2.0/validation.json
T

356 lines
22 KiB
JSON

{
"data": [
{
"caption": "00",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000003/batch2_split_8_000026_Bravo_instance_001_clip001701.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000003/batch2_split_8_000026_Bravo_instance_001_clip001701.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000003/batch2_split_8_000026_Bravo_instance_001_clip001701.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_81f_test/000000/75_batch2_split_2_000068_Bravo_instance_003_clip001053.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000000/75_batch2_split_2_000068_Bravo_instance_003_clip001053.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000000/75_batch2_split_2_000068_Bravo_instance_003_clip001053.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_81f_test/000000/75_batch2_split_1_000039_Alpha_instance_000_clip002511.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000000/75_batch2_split_1_000039_Alpha_instance_000_clip002511.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000000/75_batch2_split_1_000039_Alpha_instance_000_clip002511.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_81f_test/000001/batch2_split_14_000059_Bravo_instance_002_clip000405.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000001/batch2_split_14_000059_Bravo_instance_002_clip000405.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000001/batch2_split_14_000059_Bravo_instance_002_clip000405.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_81f_test/000001/batch2_split_10_000057_Bravo_instance_002_clip000081.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000001/batch2_split_10_000057_Bravo_instance_002_clip000081.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000001/batch2_split_10_000057_Bravo_instance_002_clip000081.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_81f_test/000001/batch2_split_0_000020_Bravo_instance_002_clip000162.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000001/batch2_split_0_000020_Bravo_instance_002_clip000162.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000001/batch2_split_0_000020_Bravo_instance_002_clip000162.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_81f_test/000000/75_batch2_split_3_000026_Bravo_instance_001_clip001863.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000000/75_batch2_split_3_000026_Bravo_instance_001_clip001863.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000000/75_batch2_split_3_000026_Bravo_instance_001_clip001863.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_81f_test/000000/75_batch2_split_2_000068_Alpha_instance_003_clip001053.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000000/75_batch2_split_2_000068_Alpha_instance_003_clip001053.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000000/75_batch2_split_2_000068_Alpha_instance_003_clip001053.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_81f_test/000003/batch2_split_9_000041_Alpha_instance_000_clip000810.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000003/batch2_split_9_000041_Alpha_instance_000_clip000810.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000003/batch2_split_9_000041_Alpha_instance_000_clip000810.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_81f_test/000002/batch2_split_4_000076_Bravo_instance_001_clip000000.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000002/batch2_split_4_000076_Bravo_instance_001_clip000000.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000002/batch2_split_4_000076_Bravo_instance_001_clip000000.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_81f_test/000000/75_batch2_split_2_000041_Bravo_instance_003_clip002673.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000000/75_batch2_split_2_000041_Bravo_instance_003_clip002673.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000000/75_batch2_split_2_000041_Bravo_instance_003_clip002673.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_81f_test/000002/batch2_split_5_000053_Alpha_instance_003_clip000243.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000002/batch2_split_5_000053_Alpha_instance_003_clip000243.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000002/batch2_split_5_000053_Alpha_instance_003_clip000243.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_81f_test/000002/batch2_split_19_000080_Alpha_instance_002_clip001296.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000002/batch2_split_19_000080_Alpha_instance_002_clip001296.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000002/batch2_split_19_000080_Alpha_instance_002_clip001296.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_81f_test/000000/75_batch2_split_1_000039_Bravo_instance_000_clip000486.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000000/75_batch2_split_1_000039_Bravo_instance_000_clip000486.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000000/75_batch2_split_1_000039_Bravo_instance_000_clip000486.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_81f_test/000000/75_batch2_split_1_000039_Bravo_instance_000_clip000162.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000000/75_batch2_split_1_000039_Bravo_instance_000_clip000162.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000000/75_batch2_split_1_000039_Bravo_instance_000_clip000162.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_81f_test/000000/75_batch2_split_2_000052_Bravo_instance_000_clip000162.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000000/75_batch2_split_2_000052_Bravo_instance_000_clip000162.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000000/75_batch2_split_2_000052_Bravo_instance_000_clip000162.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_81f_test/000001/batch2_split_0_000020_Alpha_instance_002_clip000324.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000001/batch2_split_0_000020_Alpha_instance_002_clip000324.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000001/batch2_split_0_000020_Alpha_instance_002_clip000324.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_81f_test/000001/batch2_split_10_000002_Alpha_instance_000_clip000243.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000001/batch2_split_10_000002_Alpha_instance_000_clip000243.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000001/batch2_split_10_000002_Alpha_instance_000_clip000243.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_81f_test/000002/batch2_split_2_000098_Alpha_instance_000_clip000162.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000002/batch2_split_2_000098_Alpha_instance_000_clip000162.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000002/batch2_split_2_000098_Alpha_instance_000_clip000162.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_81f_test/000002/batch2_split_6_000011_Alpha_instance_003_clip000000.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000002/batch2_split_6_000011_Alpha_instance_003_clip000000.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000002/batch2_split_6_000011_Alpha_instance_003_clip000000.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_81f_test/000000/75_batch2_split_1_000039_Alpha_instance_000_clip002754.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000000/75_batch2_split_1_000039_Alpha_instance_000_clip002754.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000000/75_batch2_split_1_000039_Alpha_instance_000_clip002754.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_81f_test/000002/batch2_split_4_000076_Bravo_instance_001_clip002673.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000002/batch2_split_4_000076_Bravo_instance_001_clip002673.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000002/batch2_split_4_000076_Bravo_instance_001_clip002673.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_81f_test/000001/75_batch2_split_8_000039_Bravo_instance_000_clip001296.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000001/75_batch2_split_8_000039_Bravo_instance_000_clip001296.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000001/75_batch2_split_8_000039_Bravo_instance_000_clip001296.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_81f_test/000003/batch2_split_8_000051_Alpha_instance_001_clip000891.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000003/batch2_split_8_000051_Alpha_instance_001_clip000891.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000003/batch2_split_8_000051_Alpha_instance_001_clip000891.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_81f_test/000002/batch2_split_19_000080_Alpha_instance_002_clip000891.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000002/batch2_split_19_000080_Alpha_instance_002_clip000891.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000002/batch2_split_19_000080_Alpha_instance_002_clip000891.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_81f_test/000001/batch2_split_0_000020_Alpha_instance_002_clip000648.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000001/batch2_split_0_000020_Alpha_instance_002_clip000648.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000001/batch2_split_0_000020_Alpha_instance_002_clip000648.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_81f_test/000002/batch2_split_1_000098_Alpha_instance_000_clip000081.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000002/batch2_split_1_000098_Alpha_instance_000_clip000081.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000002/batch2_split_1_000098_Alpha_instance_000_clip000081.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_81f_test/000002/batch2_split_5_000053_Alpha_instance_003_clip000000.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000002/batch2_split_5_000053_Alpha_instance_003_clip000000.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000002/batch2_split_5_000053_Alpha_instance_003_clip000000.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_81f_test/000001/batch2_split_16_000029_Alpha_instance_003_clip000162.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000001/batch2_split_16_000029_Alpha_instance_003_clip000162.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000001/batch2_split_16_000029_Alpha_instance_003_clip000162.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_81f_test/000000/75_batch2_split_0_000015_Bravo_instance_000_clip000486.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000000/75_batch2_split_0_000015_Bravo_instance_000_clip000486.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000000/75_batch2_split_0_000015_Bravo_instance_000_clip000486.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_81f_test/000000/75_batch2_split_4_000042_Bravo_instance_003_clip000081.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000000/75_batch2_split_4_000042_Bravo_instance_003_clip000081.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000000/75_batch2_split_4_000042_Bravo_instance_003_clip000081.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_81f_test/000002/batch2_split_19_000080_Bravo_instance_002_clip000000.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000002/batch2_split_19_000080_Bravo_instance_002_clip000000.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/train_81f_test/000002/batch2_split_19_000080_Bravo_instance_002_clip000000.npy",
"video_path": null,
"num_inference_steps": 50,
"height": 352,
"width": 640,
"num_frames": 81
}
]
}