github/MP-SPDZ
1
1
Fork 0
mirror of https://github.com/data61/MP-SPDZ.git synced 2026-05-13 03:00:24 -04:00
Code Issues Packages Projects Releases Wiki Activity
993 Commits 2 Branches 42 Tags
master
Commit Graph

449 Commits

Author SHA1 Message Date
Marcel Keller
71631ce9a4 Fix bug in binary memory access. 2026-03-12 14:02:20 +11:00
Marcel Keller
029d799ba2 Use budget with CISC instructions. 2026-03-10 13:45:46 +11:00
Vincent Ehrmanntraut
fb5271423a Merge branch 'data61:master' into master 2026-03-04 09:55:39 +01:00
Vincent Ehrmanntraut
12bc9b5330 Fix quadratic runtime of sint.concat 2026-03-04 09:52:36 +01:00
Marcel Keller
2ee81da730 Merge pull request #1768 from pps-lab/bert_layer_improvements_pr 
Reduced compilation time of BERT and other improvements
 2026-02-17 11:51:35 +11:00
Hidde L
7b72a46fe2 - Add layers like Gelu and LayerNorm explicitly to layers_from_torch so they are more generally usable (beyond BertLayer) 
- Fix Dense.__repr__ to include the missing d (sequence length) parameter
- Default LayerNorm and BertLayer to use approximate rsqrt (approx=True) for better MPC performance
- Remove debug prints
- Fix SubMultiArray.__add__ to handle addition with plain arrays (check for sizes/size attributes)
- Fix SubMultiArray.__str__ to use self.address instead of self.array._address to avoid attribute errors
- Add LAYER_COMPARISON flag to bert_inference.mpc to skip the expensive layer-by-layer comparison section (~95% of compile time) by default
 2026-02-13 09:00:51 +01:00
Marcel Keller
bf7f8f4b65 Expected communication cost in compiler. 2025-12-24 13:47:42 +11:00
Marcel Keller
f10864f85e Merge pull request #1750 from Jambon8/master 
Update types.py
 2025-12-16 12:37:38 +11:00
Marcel Keller
ce83a3708c Fix bug in exponentiation. 2025-12-15 12:40:41 +11:00
Jamil
fa4a343c4d Update types.py 
Add n_threads support to Matrix sorting
 2025-12-10 09:27:24 +01:00
Marcel Keller
70fb30c4d5 Fix bug in fixed-point input in binary circuits. 2025-12-04 15:10:06 +11:00
Marcel Keller
fa49097233 Write cgf2n to socket by converting to regint. 2025-11-25 16:59:24 +11:00
Marcel Keller
4dd834cf46 Functionality to reveal data structure parts by party to socket. 2025-11-20 11:02:01 +11:00
Marcel Keller
a2ab47642b Fix regression. 2025-11-07 15:07:54 +11:00
Marcel Keller
842491b052 Merge pull request #1736 from pps-lab/bert 
Add BERT to ML Library
 2025-11-07 15:06:37 +11:00
Hidde L
bfe3e7e038 Make interface of Dropout backwards compatible 2025-11-06 16:16:21 -05:00
Marcel Keller
383c18f1cb Fix bug with standard budget. 2025-10-30 09:35:44 +11:00
Hidde L
e17e175b16 Merge FlexDropout into Dropout 
Update mnist_full examples using Dropout
Clean up code style
 2025-10-21 16:26:15 +02:00
Hidde L
ba2b440712 Merge FlexDense with Dense 2025-10-21 15:55:20 +02:00
Hidde L
9ac22e3650 fix flexdense 2025-10-21 15:45:07 +02:00
Hidde L
2162b79b73 Revert unrelated changes 2025-10-20 11:14:38 +02:00
Marcel Keller
7e576fab0d Fix bug with small budget. 2025-10-16 12:29:22 +11:00
Hidde L
d0f955c2d3 avg diff, dropout 2025-10-14 21:43:18 +02:00
Hidde L
f8139db805 Add BERT layers and test script 2025-10-14 19:58:44 +02:00
Marcel Keller
1986dce374 Fix bug in memory access of cleartext fixed-point values with binary circuits. 2025-10-08 12:08:10 +11:00
Marcel Keller
a0c642db5d Fix bug in fixed-point addition with binary circuits. 2025-10-07 12:20:45 +11:00
Marcel Keller
bc99d71db9 Fix bug with arithmetic operations on comparison results in binary circuits. 2025-10-07 12:19:47 +11:00
Marcel Keller
dca07846e5 Fix bug in binary vector assignment. 2025-10-07 12:08:47 +11:00
Marcel Keller
dacb008625 Fix bug in optimized loop. 2025-10-03 09:35:46 +10:00
Marcel Keller
5c8884b681 Fix bug in sigmoid with binary circuits. 2025-10-02 11:57:08 +10:00
Marcel Keller
cc6b0b9b44 Fix bug when comparing bits in binary circuit. 2025-09-30 14:38:03 +10:00
Marcel Keller
24baa6d4bd Fix bug in sfix-sint operations with binary circuits. 2025-09-30 14:37:40 +10:00
Marcel Keller
9b8270fb52 Fix bug with recursive branching. 2025-09-18 10:58:41 +08:00
Marcel Keller
9cb7c40eaf Fix bug in binary register update. 2025-09-17 16:34:59 +08:00
Marcel Keller
d6049eef8f Fix bug in fixed-point division with constant-round protocol. 2025-09-17 16:30:08 +08:00
Marcel Keller
414075421b Fix bug in conversion from regint to sfix in binary. 2025-09-17 16:25:46 +08:00
Marcel Keller
e18f875f9c Fix summation of binary vectors. 2025-09-17 16:18:24 +08:00
Marcel Keller
99c63a6fbe Sample masking in linear regression. 2025-09-12 17:00:30 +08:00
Marcel Keller
b52d0ac5da Sample masking in logistic regression. 2025-09-11 17:01:44 +08:00
Marcel Keller
7fb703efe1 More precise range for NaN in fixed-point output. 2025-09-10 11:36:51 +08:00
Marcel Keller
bbf9132016 More predictable preprocessing in full bit decomposition. 2025-08-26 17:26:33 +10:00
MozirDmitriy
3fd9c20e3e Update floatingpoint.py 2025-08-11 15:55:16 +03:00
MozirDmitriy
2410571c40 fix typos 2025-08-11 15:53:36 +03:00
Marcel Keller
cf51528de6 Missing operations in binary computation. 2025-08-05 11:51:44 +02:00
crStiv
e88254f1d9 Update compilerLib.py 2025-07-26 21:14:49 +02:00
crStiv
26abe35dd3 Update comparison.py 2025-07-26 21:13:33 +02:00
crStiv
e8312b980c Update types.py 2025-07-26 21:08:11 +02:00
Marcel Keller
1d7496661b Secret sharing output for sfix and sfloat. 2025-06-06 20:08:39 +10:00
Marcel Keller
81d3b000dd Missing operator overloading. 2025-06-04 18:04:53 +10:00
Marcel Keller
9f2a641a94 Protocols with function-dependent preprocessing. 2025-05-30 13:35:02 +10:00