github/concrete
1
1
Fork 0
mirror of https://github.com/zama-ai/concrete.git synced 2026-02-09 03:55:04 -05:00
Code Issues Packages Projects Releases Wiki Activity
476 Commits 185 Branches 425 Tags
2fa3a8bcbee8f0414ec119520d47378dddcb2861
Commit Graph

91 Commits

Author SHA1 Message Date
Arthur Meyre
212dc36382 feat: emit loguru warning with reason for subgraph not fusing 
- catches cases with more than one variable input
- catches cases where the shapes are not the same in intermediate nodes

refs #645
 2021-10-28 14:36:12 +02:00
Umut
eedbe0606b feat(tracing): implement tracing of matmul 2021-10-27 17:11:19 +03:00
Arthur Meyre
ecfde7b233 refactor(debugging): accept several highlights per node when printing 
refs #645
 2021-10-26 17:19:33 +02:00
Benoit Chevallier-Mames
9459675cfb feat: manage signed constants 
closes #688
closes #612
 2021-10-26 10:06:15 +02:00
Arthur Meyre
624143106f refactor(compilation): remove unnecessary check in compile.py 
refs #645
 2021-10-25 17:02:43 +02:00
Umut
65af96253b feat(tracing): implement tracing of constant indexing 2021-10-25 13:14:29 +03:00
Umut
be453394fb fix(compilation): raise the appropriate error for intermediate signed integers 2021-10-20 17:15:52 +03:00
Benoit Chevallier-Mames
7bf2f09615 feat: remove support for np.invert 
remove support for np.invert and propose to the user to use bitwise_xor instead, because of impossibilities with float fusing
closes #658
 2021-10-19 10:40:23 +02:00
Arthur Meyre
384026364e test: create default_compilation_configuration fixture 
- update test code and use it where appropriate
- remove duplicate tests that lacked correctness verification
 2021-10-19 08:51:35 +02:00
Arthur Meyre
82688206f7 refactor(mlir): generate tables before converting nodes to MLIR 
- MLIRConverter becomes an abstract base class
- pass the needed informations in an extra dict to MLIR converters
- NPMLIRConverter handles the specifics of numpy tables generation
 2021-10-18 18:39:27 +02:00
Arthur Meyre
bc2ae7be47 chore: disable pylint line too long in test_compile where necessary 
- fix pylint targets for non package dirs
 2021-10-15 16:02:20 +02:00
Arthur Meyre
93e39e58f7 fix(representation): handle failure in UnivariateFunction get_table 
- if only some values are problematic, flood fill the resulting table with
other valid values
- if no valid value was generated an AssertionError will be thrown
 2021-10-15 14:17:11 +02:00
Umut
4c9c49ecd2 feat(debugging): provide a way for highlighting nodes with custom messages during printing 2021-10-15 13:06:36 +03:00
Arthur Meyre
95c48a419c refactor(tracing): preparatory work for tensor table generation 
- removed underlying_type_constructor from BaseDataType as it was scalar
specific and put it in values
- update inputset_eval to keep a sample of intermediate node values
- allows to get the proper value constructor to be used in
UnivariateFunction get_table and have tensors as inputs
 2021-10-14 10:53:21 +02:00
Benoit Chevallier-Mames
67a9bf12ca test: make that the tested function has integer inputs 
functions that we compile are supposed to have integer inputs and outputs
 2021-10-13 15:09:36 +02:00
Benoit Chevallier-Mames
6e79c0baf5 test(ci): more random inputs 
more random-looking inputs in subtest_fuse_float_binary_operations_correctness and
subtest_fuse_float_unary_operations_correctness
closes #547
 2021-10-13 15:09:36 +02:00
Arthur Meyre
a4da3b8210 feat(tracing): add output_idx information in edges 
- renamed output_index to output_idx in BaseTracer
- update tracing and fusing code to manage output_idx correctly
- update OPGraph evaluate and update_values_with_bounds to manage
output_idx
- update tests checking graph validity to have output_idx set properly
- the support of actual multi-output nodes is in #81
 2021-10-13 11:14:31 +02:00
Arthur Meyre
00916bcfdb refactor: rename ArbitraryFunction to UnivariateFunction 
- the naming has always been confusing and recent changes to the code make
this rename necessary for things to be clearer
 2021-10-11 11:36:35 +02:00
Arthur Meyre
44016cc80c feat: manage float fusing for tensors 
- add op_attributes on ArbitraryFunction
- this works if all constants are smaller than the input tensor
- otherwise it requires more advanced code and a concatenate operator which
currently does not exist
 2021-10-11 10:42:20 +02:00
Arthur Meyre
fb9cc79128 chore: make sure xfail test cases are strict 2021-10-11 10:42:20 +02:00
Benoit Chevallier-Mames
e8114cc470 feat: add management of boolean binary operators with a const scalar 
refs #126
refs #529
 2021-10-08 15:31:04 +02:00
Umut
6affa54473 feat(configuration): add option to treat warnings as errors 2021-10-07 18:40:33 +03:00
Benoit Chevallier-Mames
2da3895f1a feat: add more bivariate operations 
which are supported if one of the operands is a constant
refs #126
 2021-10-07 16:14:07 +02:00
Umut
5fce0d2920 feat(compilation): implement MLIR conversion of constant arrays 2021-10-07 16:17:48 +03:00
Benoit Chevallier-Mames
56e0ed4a11 feat: manage binary op where one input is constant 
feat #126
 2021-10-06 17:53:52 +02:00
Umut
83ea485fe1 feat(tracing): enable implicit broadcasting for binary operations 2021-10-06 10:33:12 +03:00
Umut
003bad581a feat(fhe_circuit): create FHECircuit class to combine operation graph and compiler engine 2021-10-04 12:23:37 +03:00
Arthur Meyre
a9d44f4758 feat(float_fusing): restrict to scalars before supporting tensors 2021-09-30 10:57:00 +02:00
Umut
c47dac833b refactor: replace scalars with () shaped tensors, disable python list support in inputset 2021-09-29 17:55:58 +03:00
Umut
36d732b0ae refactor: rename 'data_type' field of 'BaseValue' to 'dtype' 2021-09-29 11:51:08 +03:00
Umut
0061e01d62 feat: implement checking coherence between inputset and parameters 2021-09-24 13:49:54 +02:00
Arthur Meyre
eaf8cfb933 tests: add function for ArbitraryFunction arbitrary_func equivalence 
- this is not perfect but pretty close to the best we can do
 2021-09-23 16:57:12 +02:00
Umut
e4a06116ec fix: inputset of test_mlir_converter for dot operation 2021-09-23 16:31:52 +03:00
Arthur Meyre
86052fa43d chore(tools): update command line for isort to follow the 100 line length 2021-09-23 14:00:07 +02:00
Umut
441c4f9e7d feat: check inputset size 2021-09-20 17:05:22 +03:00
Arthur Meyre
381c81b76c refacto: remove iter usage for passing a dataset 
- it is still supported but not required and more confusing
 2021-09-15 20:07:21 +02:00
youben11
bd8dca11d5 fix(TLU): extend TLU to 2 ** bit_width elements 2021-09-14 17:57:52 +01:00
Benoit Chevallier-Mames
8522e58280 refacto: rename 'dataset' into a clear 'inputset' 
closes #340
 2021-09-14 17:31:55 +02:00
Benoit Chevallier-Mames
e78086eefa feat: let's have a customizable assert 
closes #245
 2021-09-10 17:52:09 +02:00
Arthur Meyre
959328e0f5 chore: rename hdk to concrete first pass 2021-09-06 18:43:49 +02:00
youben11
31c1787af2 feat(mlir): conversion of dot node into MLIR 2021-09-03 08:57:07 +01:00
youben11
f686ca535a feat(mlir): convert TensorValue inputs to MLIR 
factorized the input type conversion of scalar values
 2021-09-03 08:57:07 +01:00
Benoit Chevallier-Mames
cfe48cca15 test: add correctness tests for all supported ufunc's 
closes #263
 2021-09-02 16:00:33 +02:00
Arthur Meyre
bf2585ba0a refacto: rename hnumpy to numpy as requested for the package imports 2021-09-02 12:20:01 +02:00
Umut
1e8debfb57 refactor: rename ClearValue/EncryptedValue to ClearScalar/EncryptedScalar 2021-09-01 15:33:19 +03:00
Umut
1fa049d914 feat: dump compilation artifacts automatically on compilation failures 2021-08-31 18:40:40 +03:00
Umut
02fbbfeaf7 test(drawing): crate test for saving the drawing 2021-08-31 18:40:40 +03:00
youben11
784158741e feat(mlir): TLU conversion 2021-08-31 09:04:24 +01:00
Arthur Meyre
dbda93639b dev(ir): add get_table function to ArbitraryFunction node 2021-08-31 09:04:24 +01:00
Arthur Meyre
1b489b281c refactor: move is_equivalent to conftest.py for tests 2021-08-26 14:15:30 +02:00