Home › Census › gumbel_softmax
gumbel_softmax
gumbel_softmax: 19 implementations from 20 papers ran on one shared input (census 2026-09-22, battery b986f7e04d79); they produced 12 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:
- Every examined implementation of this name that took an array argument ran on the shared input.
- 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
18 implementations from 19 papers share this bucket (rank, kind, dtype of each array argument, positional; each member's recorded signature, argument name included, is shown under it); 11 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 |
|---|---|---|
| 4 implementations 4 papers 4b71b65a282d recorded values differ by up to 1.86e-08 |
models_dy.py 8a5f974f recorded logits:2/float/float32; temperature=0.5, hard=False model/pytorch/model.py a71884d3 recorded logits:2/float/float32; temperature=0.5, hard=False, eps=1e-10 model_vmtl.py ad6724b2 recorded logits:2/float/float32; temperature=0.5, hard=False gumbel_sigmoid_softmax.py ae5d3a40 recorded logits:2/float/float32; temperature=0.5, hard=False |
[0.220078, 0.00196555, 0.00129395, 0.000179896, 0.000346918, 0.00180161, 0.00153395, 0.7728, …] shape [4, 8] · float32 · Tensor |
| 3 implementations 4 papers ba691d412b40 recorded values identical |
one code sha held from 2 papers' repositories vqapc_model.py 4763872b recorded logits:2/float/float32; temperature=0.5
models/LSINet.py 447344e5 recorded logits:2/float/float32; temperature=0.5, hard=True, eps=1e-10, device=None model/cross_modal_attention.py c8519676 recorded logits:2/float/float32; dim=1, tau=1.0 |
[0, 0, 0, 0, 0, 0, 0, 1, …] shape [4, 8] · float32 · Tensor |
| 2 implementations 2 papers 2ddd2140e02e recorded values identical |
maddpg-pytorch/algorithms/maddpg.py 57d033c0 recorded logits:2/float/float32; temperature=1.0, hard=False gumbel_sigmoid_softmax.py 706773b4 recorded logits:2/float/float32; temperature=1.0, hard=False |
[0.304198, 0.0287482, 0.0233253, 0.00869719, 0.0120776, 0.0275231, 0.0253964, 0.570035, …] shape [4, 8] · float32 · Tensor |
| 2 implementations 2 papers 8c514d142782 recorded values identical |
model_search.py 1824fbfc recorded logits:2/float/float32; temperature=1.0, hard=False model_search.py 51090852 recorded logits:2/float/float32; temperature=1.0, hard=False |
[0.304198, 0.0287482, 0.0233253, 0.00869719, 0.0120776, 0.0275231, 0.0253964, 0.570035, …] shape [4, 8] · float32 · Tensor |
| 1 implementation 2 papers d518dad29681 |
model.py 7f291425 recorded logits:2/float/float32; tau=1.0, hard=False, dim=-1 |
[0.304198, 0.0287482, 0.0233253, 0.00869719, 0.0120776, 0.0275232, 0.0253965, 0.570034, …] shape [4, 8] · float32 · Tensor |
| 1 implementation 1 paper 08b19d1cfcec |
model/RGSL.py 7e5297fa recorded logits:2/float/float32; tau=1.0, hard=False, eps=1e-10, dim=-1 |
[0.54749, -1.7411, 0.962115, -1.62113, -0.937932, -1.47908, 0.963047, 1.41109, …] shape [4, 8] · float32 · Tensor |
| 1 implementation 1 paper 3fcfa9b79d41 |
mixed_operation.py 3cb0d3bd recorded alphas:2/float/float32; temp=0.5, epsilon=0.0001 |
[0.193988, 4.21537e-09, 0.00112944, 7.14905e-11, 2.1065e-10, 1.41204e-09, 5.76245e-10, 0.804883, …] shape [4, 8] · float32 · Tensor |
| 1 implementation 1 paper 5a943745493c |
src/models/dams.py e6d0d3fd recorded logits:2/float/float32; tau=1.0, hard=False, log_mode=True, dim=-1 |
[-1.94228, -4.23087, -1.52766, -4.1109, -3.4277, -3.96885, -1.52672, -1.07868, …] shape [4, 8] · float32 · Tensor |
| 1 implementation 1 paper 8de8fd95c344 |
model/GroupNet_nba.py 422ac583 recorded logits:2/float/float32; tau=1.0, hard=False, eps=1e-10 |
[0.60614, 0.0818282, 0.0120068, 0.134808, 0.105253, 0.0575458, 0.349442, 0.348225, …] shape [4, 8] · float32 · Tensor |
| 1 implementation 1 paper 9afd589f4243 |
gnia.py 36c38bfa recorded logits:2/float/float32; tau=0.5, random_flag=True, eps=0.1, dim=-1 |
[0.832682, 0.00128057, 0.0457499, 0.00459693, 0.0033519, 0.003141, 0.00599147, 0.103206, …] shape [4, 8] · float32 · Tensor |
| 1 implementation 1 paper f2dd12d82e68 |
cnn/model_search.py d599e623 recorded logits:2/float/float32; tau=1.0, hard=False, eps=1e-10, dim=-1 |
[0.143377, 0.0145397, 0.217044, 0.0163931, 0.0324614, 0.0188952, 0.217246, 0.340044, …] shape [4, 8] · float32 · Tensor |
Bucket 2 of 2: arg 1: rank 2, kind float, dtype float32 · arg 2: rank 2, 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 1d026df1ea20 |
models/sparsefunc.py ef021918 recorded logits:2/float/float32, U:2/float/float32; temperature=0.5, hard=False, eps=1e-20 |
[nan, nan, 0.681437, nan, nan, nan, nan, 0.849805, …] shape [4, 8] · float32 · Tensor · non-finite |
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/gumbel-softmax.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