sinkhorn
sinkhorn: 11 implementations from 11 papers ran on one shared input (census 2026-09-22, battery b986f7e04d79); they produced 11 distinct outputs across 2 buckets, one shared input per bucket.
Identical values to six decimals (the recorded digest) on the shared input are agreement on those inputs, not equivalence. Implementations are compared only within one bucket, the positional (rank, kind, dtype) of each array argument; the argument name is not part of the key because the harness draws the shared array from (rank, kind) and casts it to the dtype, whatever the name; each bucket's shared input is fixed by that key, so members of one bucket saw bitwise-identical inputs under their own scalar arguments. A cluster is the set of members whose recorded output digest is identical. Nothing here says which computation a paper's method intended, and nothing reproduces a paper's results.
Not compared, and not in the tables or the counts above:
- 9 more implementations of this name were examined but did not run on the shared input (RuntimeError 4, ValueError 3, AssertionError 1, IndexError 1).
- No implementation of this name ran on its own fixture arguments only: every one that ran took an array argument.
- Every compared output was digested.
Bucket 1 of 2: arg 1: rank 2, kind float, dtype float32
10 implementations from 10 papers share this bucket (rank, kind, dtype of each array argument, positional; each member's recorded signature, argument name included, is shown under it); 10 distinct outputs, largest cluster first. Values are the first 8 of the recorded output, flattened.
| Cluster (same digest to six decimals) | Members | Shared output on this bucket's input |
|---|---|---|
| 1 implementation 1 paper 51e63d44301d |
lightly/loss/msn_loss.py c1dff37f recorded probabilities:2/float/float32; iterations=3, gather_distributed=False |
[0.0939011, -0.159042, -0.0528384, 0.0394712, 0.975307, -0.0744611, -0.151851, 0.329514, …] shape [4, 8] · float32 · Tensor |
| 1 implementation 1 paper 6b36e4973088 |
model.py ede20b10 recorded scores:2/float/float32; eps=0.05, niters=3 |
[0.86799, 1.15486e-21, 0.00026542, 4.55076e-15, 9.98209e-08, 3.94145e-29, 0.130547, 0.00119711, …] shape [4, 8] · float32 · Tensor |
| 1 implementation 1 paper 93d803bcd448 |
time_tuning.py 0988ed56 recorded Q:2/float/float32; nmb_iters=2, world_size=1 |
[20.174, -6.27716, 0.594783, -13.4916, 0.477349, 0.322681, 0.404291, -0.204321, …] shape [8, 4] · float32 · Tensor |
| 1 implementation 1 paper 96f9141d2b84 |
models/layers.py 3e002910 recorded out:2/float/float32; epsilon=0.05, sinkhorn_iterations=3 |
[1, 1.04799e-29, 3.00271e-12, 2.19583e-22, 3.16604e-24, 2.46609e-25, 3.85389e-22, 2.08267e-11, …] shape [4, 8] · float32 · Tensor |
| 1 implementation 1 paper ae8daa487fdd |
switch_mlp.py bfbd1627 recorded cost:2/float/float32; tol=0.0001 |
[0.11719, 0.00123529, 0.0336215, 0.00522136, 0.014975, 0.000292158, 0.0418428, 0.0344142, …] shape [4, 8] · float32 · Tensor |
| 1 implementation 1 paper b1ec41a78738 |
experiments/exp_mosic.py 432bf71f recorded Q:2/float/float32; nmb_iters=5, world_size=1 |
[0.533104, 0.494034, 0.247564, -0.274702, 0.0833526, -0.167814, 1.11195, -0.0274899, …] shape [8, 4] · float32 · Tensor |
| 1 implementation 1 paper da8384339a92 |
model_symmetries/stitching/stitching.py 57f3995b recorded A:2/float/float32; iters=16, verbose=False |
[-2.99491, 0.217749, 0.12607, -0.161571, 0.50492, 0.123525, 0.314265, 2.1925, …] shape [4, 8] · float32 · Tensor |
| 1 implementation 1 paper f6fc4ba3d136 |
lightly/loss/swav_loss.py 58db645a recorded out:2/float/float32; iterations=3, epsilon=0.05, gather_distributed=False |
[0.867991, 1.15486e-21, 0.00026542, 4.55076e-15, 9.98207e-08, 3.94145e-29, 0.130547, 0.00119711, …] shape [4, 8] · float32 · Tensor |
| 1 implementation 1 paper f973c3905903 |
methods/uniot.py 2be40bf6 recorded out:2/float/float32; epsilon=0.05, sinkhorn_iterations=3 |
[0.216998, 2.88716e-22, 6.6355e-05, 1.13769e-15, 2.49552e-08, 9.85362e-30, 0.0326367, 0.000299279, …] shape [4, 8] · float32 · Tensor |
| 1 implementation 1 paper fadf10afaef3 |
loss.py 8cf6df4a recorded M:2/float/float32; reg=10, numItermax=10000 |
[-0.165794] shape [] · float32 · Tensor |
Bucket 2 of 2: arg 1: rank 3, kind float, dtype float32
1 implementation from 1 paper share this bucket (rank, kind, dtype of each array argument, positional; each member's recorded signature, argument name included, is shown under it); 1 distinct output, largest cluster first. Values are the first 8 of the recorded output, flattened.
| Cluster (same digest to six decimals) | Members | Shared output on this bucket's input |
|---|---|---|
| 1 implementation 1 paper 0d678eb6f456 |
lib/learnable_sparsity.py 78199858 recorded logits:3/float/float32; tau=1.0, iter=5 |
[0.0799626, 0.668082, 0.203903, 0.290885, 0.305568, 0.0526106, 0.302167, 0.0969119, …] shape [2, 4, 8] · float32 · Tensor |
Identical values to six decimals (the recorded digest) on the shared input are agreement on those inputs, not equivalence; where a cluster's members carry recorded values, the largest difference among them is shown under the cluster. Paper titles are the archive's archive 2025-07-28 where the paper is in the archive and the graph's where it was added by Syntology; papers with no page here are shown by their recorded paper id only. A paper count above the implementation count means one implementation (one code sha) is held from several papers' repositories and counts once. Per-sample status, licence and fingerprint records for each paper are on its paper page. JSON twin: /census/sinkhorn.json.
Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections