vista3d-ct-mlx / reference-comparison.json
josand's picture
Upload converted VISTA3D weights and release metadata
1134220 verified
Raw History Blame Contribute Delete
18.3 kB
{
"schema_version": "medmlx.reference.v1",
"model": "vista3d",
"date": "2026-10-12T17:35:32.676541+00:00",
"case": {
"dataset": "AMOS22 (validation)",
"case_id": "amos_0008",
"input_sha256": "e3ed0f4d7b1fcd1b785ffdb233750cc35eff9714fa724b07122dd625a9c85461"
},
"original": {
"code": "NVIDIA NV-Segment-CTMR / MONAI VISTA3D vista3d132",
"version_or_commit": "f9f5f51b589e5dc9c23c453cf5138398e4084056; MONAI 1.6.0, PyTorch 2.14.1",
"precision": "float32, AMP disabled for CPU",
"command": "cd <paper-repo>/work/reference/vista3d/upstream/NV-Segment-CT && <paper-repo>/work/reference/vista3d/.venv/bin/python -m monai.bundle run run --meta_file configs/metadata.json --config_file configs/inference.json --bundle_root <paper-repo>/work/reference/vista3d/upstream/NV-Segment-CT --device cpu --checkpoint_path <home>/.cache/huggingface/hub/models--nvidia--NV-Segment-CT/snapshots/afb51518689f71e6abb367ee6301b2cd0225c66a/vista3d_pretrained_model/model.pt --input_dict '{\"image\": \"<repositories>/research/radnn/work/nnunet-next-20261001/public-ct/amos_0008/image.nii.gz\"}' --output_dir <paper-repo>/work/reference/vista3d/out '--evaluator#amp' False '--initialize#0' '$monai.utils.set_determinism(seed=123)'",
"source_files_sha256": {
"scripts/__init__.py": "0dde80e8d0dbf0b7ce3746cfb420a75ce8a108c9d969d1b2426220a071aec51a",
"scripts/batch_inference_utils.py": "52da0fcd37f7d04990c90b1b94342c65bb0ac9529569d7963f6c728b4e2c67f6",
"scripts/early_stop_score_function.py": "e5c77b011847941aedfc7aa5c0e07a72988ed676ffb0356e211aa61c24f9d7d3",
"scripts/evaluator.py": "c985aab7b08b2d58299d01f2403f3fd9cb2b49d5ea93dd16867c54d354c8d1da",
"scripts/huggingface_download.py": "498dc8816bfd451f704c54e9ad949dd4e90d66102352aed9db3c400aa18fa140",
"scripts/inferer.py": "4cf51eb5c6eae9e5c52e4da4227b7c068392ed0b286a229b13fb5df52660b6f1",
"scripts/trainer.py": "b236eb552d20672a3f72a16fdb13101ee6e8f547a5c6ec51cb73c5e51aa482fb",
"configs/batch_inference.json": "287271ff182317fc8661587328d450a6a23f6f8cdda9c1bad624e8586ad547a5",
"configs/evaluate.json": "4e91d4031fd152f2e3f7219265b7ce1fed36e84db8cc7d4bfedcd0318c0682ec",
"configs/inference.json": "fdb0fb917009521955a753c617afdd6ffae241320efbc6d05a4f3f7d97f6ae61",
"configs/inference_trt.json": "2d70154125ced59cab280599546db8e2609423e1a0fb055c7cce7ce9bb2789b4",
"configs/label_dict.json": "186226ba214b3e02cc5394427c6395364905129e449a6b726253269117302475",
"configs/metadata.json": "a90caa13baaea9c025430c0fd6d5eaae1bfb68c84826fbd095e7897832a6903b",
"configs/mgpu_evaluate.json": "9cebeaad28fdcb3526c0e2752dfae389490847a7bf768b36f5196fa93233e143",
"configs/mgpu_inference.json": "1354bf905c3deb7d3780883a7ee3d328853eadf4e51ac49513601b99f0e59956",
"configs/msd_task09_spleen_folds.json": "08607e2e345ebddf7b2ffae6ef1f4ddec6ab9272697d9cc5501354e621bcc8d6",
"configs/multi_gpu_train.json": "147ac052a94849a184fc315d32979b70d64d315a330d9a78c684b97c959c1e17",
"configs/train.json": "4117406ff7966f0c6b121106e46c59f74d4c00427560a2a73608a2cad645ca86",
"configs/train_continual.json": "6336677108895f5518386fb5dd8512981edd02854a75c877244cc84e826ca7dd"
},
"device": "cpu",
"checkpoint_sha256": "c92bab26d00b4a5d89fa8a383900cdeb88302fd318e5e816df0bbec7106d9a1b"
},
"mlx": {
"run_record": "results/runs/vista3d.json",
"output_path": "<paper-repo>/work/runs/vista3d/out/nv_segment_ct_mask.nii.gz",
"output_sha256": "5e77bb9d0e17802ff231238e7057887fa295c35a230053d8468fec46d75bdc51"
},
"comparison": {
"metric": "minimum per-label mask Dice",
"value": 1.0,
"per_label": {
"0": {
"dice": 1.0,
"original_voxels": 54160900,
"mlx_voxels": 54160900,
"intersection_voxels": 54160900
},
"1": {
"dice": 1.0,
"original_voxels": 497534,
"mlx_voxels": 497534,
"intersection_voxels": 497534
},
"3": {
"dice": 1.0,
"original_voxels": 47788,
"mlx_voxels": 47788,
"intersection_voxels": 47788
},
"4": {
"dice": 1.0,
"original_voxels": 46689,
"mlx_voxels": 46689,
"intersection_voxels": 46689
},
"5": {
"dice": 1.0,
"original_voxels": 62218,
"mlx_voxels": 62218,
"intersection_voxels": 62218
},
"6": {
"dice": 1.0,
"original_voxels": 49836,
"mlx_voxels": 49836,
"intersection_voxels": 49836
},
"7": {
"dice": 1.0,
"original_voxels": 29191,
"mlx_voxels": 29191,
"intersection_voxels": 29191
},
"8": {
"dice": 1.0,
"original_voxels": 2022,
"mlx_voxels": 2022,
"intersection_voxels": 2022
},
"9": {
"dice": 1.0,
"original_voxels": 2064,
"mlx_voxels": 2064,
"intersection_voxels": 2064
},
"10": {
"dice": 1.0,
"original_voxels": 8568,
"mlx_voxels": 8568,
"intersection_voxels": 8568
},
"11": {
"dice": 1.0,
"original_voxels": 5198,
"mlx_voxels": 5198,
"intersection_voxels": 5198
},
"12": {
"dice": 1.0,
"original_voxels": 96468,
"mlx_voxels": 96468,
"intersection_voxels": 96468
},
"13": {
"dice": 1.0,
"original_voxels": 28373,
"mlx_voxels": 28373,
"intersection_voxels": 28373
},
"14": {
"dice": 1.0,
"original_voxels": 65772,
"mlx_voxels": 65772,
"intersection_voxels": 65772
},
"15": {
"dice": 1.0,
"original_voxels": 44382,
"mlx_voxels": 44382,
"intersection_voxels": 44382
},
"17": {
"dice": 1.0,
"original_voxels": 4656,
"mlx_voxels": 4656,
"intersection_voxels": 4656
},
"19": {
"dice": 1.0,
"original_voxels": 430429,
"mlx_voxels": 430429,
"intersection_voxels": 430429
},
"22": {
"dice": 1.0,
"original_voxels": 18,
"mlx_voxels": 18,
"intersection_voxels": 18
},
"28": {
"dice": 1.0,
"original_voxels": 44644,
"mlx_voxels": 44644,
"intersection_voxels": 44644
},
"29": {
"dice": 1.0,
"original_voxels": 100456,
"mlx_voxels": 100456,
"intersection_voxels": 100456
},
"30": {
"dice": 1.0,
"original_voxels": 442,
"mlx_voxels": 442,
"intersection_voxels": 442
},
"31": {
"dice": 1.0,
"original_voxels": 98823,
"mlx_voxels": 98823,
"intersection_voxels": 98823
},
"32": {
"dice": 1.0,
"original_voxels": 145317,
"mlx_voxels": 145317,
"intersection_voxels": 145317
},
"33": {
"dice": 1.0,
"original_voxels": 41023,
"mlx_voxels": 41023,
"intersection_voxels": 41023
},
"34": {
"dice": 1.0,
"original_voxels": 39950,
"mlx_voxels": 39950,
"intersection_voxels": 39950
},
"35": {
"dice": 1.0,
"original_voxels": 38224,
"mlx_voxels": 38224,
"intersection_voxels": 38224
},
"36": {
"dice": 1.0,
"original_voxels": 32207,
"mlx_voxels": 32207,
"intersection_voxels": 32207
},
"37": {
"dice": 1.0,
"original_voxels": 29374,
"mlx_voxels": 29374,
"intersection_voxels": 29374
},
"38": {
"dice": 1.0,
"original_voxels": 25004,
"mlx_voxels": 25004,
"intersection_voxels": 25004
},
"39": {
"dice": 1.0,
"original_voxels": 23054,
"mlx_voxels": 23054,
"intersection_voxels": 23054
},
"40": {
"dice": 1.0,
"original_voxels": 21771,
"mlx_voxels": 21771,
"intersection_voxels": 21771
},
"41": {
"dice": 1.0,
"original_voxels": 19435,
"mlx_voxels": 19435,
"intersection_voxels": 19435
},
"42": {
"dice": 1.0,
"original_voxels": 2145,
"mlx_voxels": 2145,
"intersection_voxels": 2145
},
"43": {
"dice": 1.0,
"original_voxels": 12,
"mlx_voxels": 12,
"intersection_voxels": 12
},
"58": {
"dice": 1.0,
"original_voxels": 10185,
"mlx_voxels": 10185,
"intersection_voxels": 10185
},
"59": {
"dice": 1.0,
"original_voxels": 8837,
"mlx_voxels": 8837,
"intersection_voxels": 8837
},
"60": {
"dice": 1.0,
"original_voxels": 19445,
"mlx_voxels": 19445,
"intersection_voxels": 19445
},
"61": {
"dice": 1.0,
"original_voxels": 15055,
"mlx_voxels": 15055,
"intersection_voxels": 15055
},
"62": {
"dice": 1.0,
"original_voxels": 340945,
"mlx_voxels": 340945,
"intersection_voxels": 340945
},
"65": {
"dice": 1.0,
"original_voxels": 8,
"mlx_voxels": 8,
"intersection_voxels": 8
},
"66": {
"dice": 1.0,
"original_voxels": 128,
"mlx_voxels": 128,
"intersection_voxels": 128
},
"67": {
"dice": 1.0,
"original_voxels": 4699,
"mlx_voxels": 4699,
"intersection_voxels": 4699
},
"68": {
"dice": 1.0,
"original_voxels": 7946,
"mlx_voxels": 7946,
"intersection_voxels": 7946
},
"69": {
"dice": 1.0,
"original_voxels": 8893,
"mlx_voxels": 8893,
"intersection_voxels": 8893
},
"70": {
"dice": 1.0,
"original_voxels": 9148,
"mlx_voxels": 9148,
"intersection_voxels": 9148
},
"71": {
"dice": 1.0,
"original_voxels": 10478,
"mlx_voxels": 10478,
"intersection_voxels": 10478
},
"72": {
"dice": 1.0,
"original_voxels": 7125,
"mlx_voxels": 7125,
"intersection_voxels": 7125
},
"73": {
"dice": 1.0,
"original_voxels": 4895,
"mlx_voxels": 4895,
"intersection_voxels": 4895
},
"74": {
"dice": 1.0,
"original_voxels": 1241,
"mlx_voxels": 1241,
"intersection_voxels": 1241
},
"78": {
"dice": 1.0,
"original_voxels": 167,
"mlx_voxels": 167,
"intersection_voxels": 167
},
"79": {
"dice": 1.0,
"original_voxels": 5002,
"mlx_voxels": 5002,
"intersection_voxels": 5002
},
"80": {
"dice": 1.0,
"original_voxels": 5959,
"mlx_voxels": 5959,
"intersection_voxels": 5959
},
"81": {
"dice": 1.0,
"original_voxels": 8899,
"mlx_voxels": 8899,
"intersection_voxels": 8899
},
"82": {
"dice": 1.0,
"original_voxels": 8954,
"mlx_voxels": 8954,
"intersection_voxels": 8954
},
"83": {
"dice": 1.0,
"original_voxels": 9927,
"mlx_voxels": 9927,
"intersection_voxels": 9927
},
"84": {
"dice": 1.0,
"original_voxels": 6338,
"mlx_voxels": 6338,
"intersection_voxels": 6338
},
"85": {
"dice": 1.0,
"original_voxels": 4899,
"mlx_voxels": 4899,
"intersection_voxels": 4899
},
"86": {
"dice": 1.0,
"original_voxels": 2011,
"mlx_voxels": 2011,
"intersection_voxels": 2011
},
"87": {
"dice": 1.0,
"original_voxels": 346,
"mlx_voxels": 346,
"intersection_voxels": 346
},
"88": {
"dice": 1.0,
"original_voxels": 407,
"mlx_voxels": 407,
"intersection_voxels": 407
},
"89": {
"dice": 1.0,
"original_voxels": 4008,
"mlx_voxels": 4008,
"intersection_voxels": 4008
},
"90": {
"dice": 1.0,
"original_voxels": 5665,
"mlx_voxels": 5665,
"intersection_voxels": 5665
},
"93": {
"dice": 1.0,
"original_voxels": 108085,
"mlx_voxels": 108085,
"intersection_voxels": 108085
},
"94": {
"dice": 1.0,
"original_voxels": 102740,
"mlx_voxels": 102740,
"intersection_voxels": 102740
},
"95": {
"dice": 1.0,
"original_voxels": 197215,
"mlx_voxels": 197215,
"intersection_voxels": 197215
},
"96": {
"dice": 1.0,
"original_voxels": 199300,
"mlx_voxels": 199300,
"intersection_voxels": 199300
},
"97": {
"dice": 1.0,
"original_voxels": 90239,
"mlx_voxels": 90239,
"intersection_voxels": 90239
},
"98": {
"dice": 1.0,
"original_voxels": 324193,
"mlx_voxels": 324193,
"intersection_voxels": 324193
},
"99": {
"dice": 1.0,
"original_voxels": 340689,
"mlx_voxels": 340689,
"intersection_voxels": 340689
},
"100": {
"dice": 1.0,
"original_voxels": 136198,
"mlx_voxels": 136198,
"intersection_voxels": 136198
},
"101": {
"dice": 1.0,
"original_voxels": 133965,
"mlx_voxels": 133965,
"intersection_voxels": 133965
},
"102": {
"dice": 1.0,
"original_voxels": 34506,
"mlx_voxels": 34506,
"intersection_voxels": 34506
},
"103": {
"dice": 1.0,
"original_voxels": 34919,
"mlx_voxels": 34919,
"intersection_voxels": 34919
},
"104": {
"dice": 1.0,
"original_voxels": 199313,
"mlx_voxels": 199313,
"intersection_voxels": 199313
},
"105": {
"dice": 1.0,
"original_voxels": 207480,
"mlx_voxels": 207480,
"intersection_voxels": 207480
},
"106": {
"dice": 1.0,
"original_voxels": 169532,
"mlx_voxels": 169532,
"intersection_voxels": 169532
},
"107": {
"dice": 1.0,
"original_voxels": 163972,
"mlx_voxels": 163972,
"intersection_voxels": 163972
},
"110": {
"dice": 1.0,
"original_voxels": 4,
"mlx_voxels": 4,
"intersection_voxels": 4
},
"113": {
"dice": 1.0,
"original_voxels": 19,
"mlx_voxels": 19,
"intersection_voxels": 19
},
"114": {
"dice": 1.0,
"original_voxels": 64939,
"mlx_voxels": 64939,
"intersection_voxels": 64939
},
"115": {
"dice": 1.0,
"original_voxels": 224622,
"mlx_voxels": 224622,
"intersection_voxels": 224622
},
"118": {
"dice": 1.0,
"original_voxels": 17342,
"mlx_voxels": 17342,
"intersection_voxels": 17342
},
"119": {
"dice": 1.0,
"original_voxels": 999,
"mlx_voxels": 999,
"intersection_voxels": 999
},
"120": {
"dice": 1.0,
"original_voxels": 150,
"mlx_voxels": 150,
"intersection_voxels": 150
},
"121": {
"dice": 1.0,
"original_voxels": 28719,
"mlx_voxels": 28719,
"intersection_voxels": 28719
},
"122": {
"dice": 1.0,
"original_voxels": 8967,
"mlx_voxels": 8967,
"intersection_voxels": 8967
},
"125": {
"dice": 1.0,
"original_voxels": 2825,
"mlx_voxels": 2825,
"intersection_voxels": 2825
},
"127": {
"dice": 1.0,
"original_voxels": 25695,
"mlx_voxels": 25695,
"intersection_voxels": 25695
}
},
"threshold": {
"min_dice": 0.99,
"scope": "every label present in either mask, including background"
},
"result": "pass",
"differing_voxels": 0,
"geometry": {
"shape": [
768,
768,
101
],
"affine_max_abs_diff": 0.0
},
"original_output_path": "<paper-repo>/work/reference/vista3d/out/image/image_trans.nii.gz",
"original_output_sha256": "5e77bb9d0e17802ff231238e7057887fa295c35a230053d8468fec46d75bdc51"
},
"notes": "Automatic segment_all: no label or point prompts; the original evaluator selects its 117 everything_labels. Same recorded checkpoint and AMOS input, original bundle preprocessing, 128^3 patches, sw_batch_size=1, replicate padding, automatic inferer's default overlap=0.25 and constant blending, and original postprocessing. CPU device, AMP=false, paths and fixed local checkpoint are explicit overrides. The initial download/counter operation is replaced by the same deterministic seed call because the original checkpoint is already present and hash-verified. No source files or package repos are edited. MLX's saved output is compared without rerunning MLX. The recorded radnn input path moved to GitHub/research/radnn; bytes match.",
"failure": null,
"verification": {
"inputs": {
"image": {
"path": "<repositories>/research/radnn/work/nnunet-next-20261001/public-ct/amos_0008/image.nii.gz",
"sha256": "e3ed0f4d7b1fcd1b785ffdb233750cc35eff9714fa724b07122dd625a9c85461"
}
},
"mlx_outputs": {
"mask": {
"path": "<paper-repo>/work/runs/vista3d/out/nv_segment_ct_mask.nii.gz",
"sha256": "5e77bb9d0e17802ff231238e7057887fa295c35a230053d8468fec46d75bdc51"
}
},
"checkpoint": {
"path": "<home>/.cache/huggingface/hub/models--nvidia--NV-Segment-CT/snapshots/afb51518689f71e6abb367ee6301b2cd0225c66a/vista3d_pretrained_model/model.pt",
"sha256": "c92bab26d00b4a5d89fa8a383900cdeb88302fd318e5e816df0bbec7106d9a1b"
}
}
}