{ "variant": "14B three encoders, checkpoint 80", "reference_count": 8000, "weights": [ 10.0, 1.0, 1.0 ], "divisor": 3.0, "batch_geometry": [ 256, 128, 128 ], "branches": { "DINOv3": { "sigma": 341.2794189453125, "landmark_shape": [ 256, 1024, 1024 ], "nystrom": { "bandwidth_source": "median_frobenius_distance", "generated_gram": "biased_exact_including_diagonal", "kernel": "gaussian_rbf", "kernel_metric": "frobenius", "kmeans": "deterministic_farthest_first_lloyd", "kmeans_iterations": 20, "median_max_samples": 2048, "num_landmarks": 256, "reference_count": 8000, "ridge": 1e-06, "sample_axis": "video", "sample_shape": [ 1024, 1024 ], "seed": 0 }, "content_hash": "67851f7828ee0b9e12643242ff93f2b2b9b7c56059e51b9b2f267972da15eba9" }, "V-JEPA 2": { "sigma": 3025.305908203125, "landmark_shape": [ 1024, 1024, 1024 ], "nystrom": { "bandwidth_source": "median_frobenius_distance", "generated_gram": "biased_exact_including_diagonal", "kernel": "gaussian_rbf", "kernel_metric": "frobenius", "kmeans": "deterministic_farthest_first_lloyd", "kmeans_iterations": 20, "median_max_samples": 2048, "num_landmarks": 1024, "reference_count": 8000, "ridge": 1e-06, "sample_axis": "video", "sample_shape": [ 1024, 1024 ], "seed": 0 }, "content_hash": "d588e77dd6ea5d5611f5aeaac2c7f9a8c4e4fc36940209df32118a91e22d168b" }, "VideoMAEv2": { "sigma": 3204.123291015625, "landmark_shape": [ 256, 1280, 1024 ], "nystrom": { "bandwidth_source": "median_frobenius_distance", "generated_gram": "biased_exact_including_diagonal", "kernel": "gaussian_rbf", "kernel_metric": "frobenius", "kmeans": "deterministic_farthest_first_lloyd", "kmeans_iterations": 20, "median_max_samples": 2048, "num_landmarks": 256, "reference_count": 8000, "ridge": 1e-06, "sample_axis": "video", "sample_shape": [ 1280, 1024 ], "seed": 0 }, "content_hash": "88d55200ebb71d20ea75c70358eb8e5e68511c030546c5f832b53e591dd5a9b2" } } }