Commits

History for

Commits on May 14, 2024

[NeoML] Allow link MKL in cross-compile (#1068 )
favorart
committedMay 14, 2024

Commits on May 10, 2024

[NeoMathEngine] GetReuseMemoryMode and GetCurrentMemoryUsage (#1065 )
favorart
committedMay 10, 2024

Commits on May 8, 2024

[NeoMathEngine] The threshold in bytes for buffer pools allocation for the thread (#1061 )
favorart
committedMay 8, 2024

Commits on Apr 25, 2024

[NeoMathEngine] fix macOS cross-compilation, disable MKL for cross-compilation (#1053 )
favorart
committedApr 25, 2024

Commits on Apr 24, 2024

[NeoMathEngine] macOS cross-compilation fix (#1052 )
favorart
committedApr 24, 2024

Commits on Apr 3, 2024

[CudaMathEngine] SetCudaDevice in copy and convert, minor CBaseLayer optimizing (#1036 )
favorart
committedApr 3, 2024

Commits on Mar 29, 2024

[CudaMathEngine] Add const modifiers (#1006 )
favorart
committedMar 29, 2024

Commits on Mar 28, 2024

Revert partial [NeoMathEngine] Add MKL SGemm JIT (#921 ) (#1041 )
favorart
committedMar 28, 2024
Android change (#1042 )

daniyalaliev
and
Valeriy Fedyunin
committedMar 28, 2024

Commits on Mar 4, 2024

[CpuMathEngine]Reduce dropout memory consumption by CPU MathEngine (#1034 )
daniyalaliev
committedMar 4, 2024

Commits on Feb 27, 2024

[NeoML] MemoryPool add ResetPeakMemoryUsage (#1032 )
favorart
committedFeb 27, 2024

Commits on Jan 18, 2024

[NeoML] CDnnBlob size or valid buffer size for TransferDataToThisThread (#1024 )

favorart
and
SAngeliuk
committedJan 18, 2024

Commits on Dec 29, 2023

[NeoML] Transfer CDnnBlob data in threads pools (#1014 )
favorart
committedDec 29, 2023

Commits on Dec 22, 2023

[CudaMathEngine] Fix shared buffers initializations and expf calls (#1004 )

favorart
and
Valeriy Fedyunin
committedDec 22, 2023

Commits on Dec 21, 2023

[CudaMathEngine] Fix restrict modifier for function arguments (#1011 )

favorart
and
Valeriy Fedyunin
committedDec 21, 2023
[CudaMathEngine] Fix warp-size iterations in max reducing functions (#1010 )
favorart
committedDec 21, 2023

Commits on Dec 20, 2023

Reduced CUDA dropout memory usage (#1005 )
daniyalaliev
committedDec 20, 2023

Commits on Dec 11, 2023

Counters linux (#1003 )

daniyalaliev
and
Valeriy Fedyunin
committedDec 11, 2023

Commits on Nov 24, 2023

removes authors list (#996 )
SanSanch5
committedNov 24, 2023

Commits on Oct 13, 2023

Fix compilation for larger UNITY_BUILD_BATCH_SIZE (#985 )
Valeriy Fedyunin
committedOct 13, 2023

Commits on Oct 6, 2023

Revert "Automatically detect memory limits (#978 )" (#981 )
Valeriy Fedyunin
committedOct 6, 2023

Commits on Oct 4, 2023

Automatically detect memory limits (#978 )
Valeriy Fedyunin
committedOct 4, 2023

Commits on Oct 3, 2023

Revert "Add memory limit to distributed training on CPU (#976 )" (#977 )
Valeriy Fedyunin
committedOct 3, 2023
Add memory limit to distributed training on CPU (#976 )
Valeriy Fedyunin
committedOct 3, 2023

Commits on Sep 27, 2023

Use both BusID and DeviceID to identify CUDA GPU (#971 )
Valeriy Fedyunin
committedSep 27, 2023

Commits on Sep 22, 2023

Fix in-place condition for rowwise MobileNetV2 block on CUDA (#969 )
Valeriy Fedyunin
committedSep 22, 2023

Commits on Sep 14, 2023

Fix tests for new Logarithm in ARM (#960 )
Valeriy Fedyunin
committedSep 14, 2023
Fix polynom in logarithm for ARM (#959 )
Valeriy Fedyunin
committedSep 14, 2023

Commits on Sep 13, 2023

Don't use custom sgemm when MLAS is available (#956 )
Valeriy Fedyunin
committedSep 13, 2023
Use erf from MKL on Intel CPUs (#958 )
Valeriy Fedyunin
committedSep 13, 2023

Commits on Sep 12, 2023

[NeoMathEngine] Add MKL SGemm JIT to some layers (#921 )
favorart
committedSep 12, 2023

Commits on Sep 2, 2023

Fix values range in MlasComputeLogistic (#954 )
Valeriy Fedyunin
committedSep 2, 2023

Commits on Aug 31, 2023

Transpose recurrent weights for faster inference on CPU (#953 )
Valeriy Fedyunin
committedAug 31, 2023
Fix sigmoid range in MLAS (#952 )
Valeriy Fedyunin
committedAug 31, 2023

Commits on Aug 30, 2023

Use MLAS vector functions (#951 )
Valeriy Fedyunin
committedAug 30, 2023