Jason Morton
0d7b6940ba
rename crate ( #39 )
...
Adjust crate name uses internally to ezkl.
2022-10-08 01:46:13 -04:00
Jason Morton
3ae716760b
fix inputs ( #37 )
...
cli now takes floats as inputs, convert examples to floats.
2022-10-07 16:07:52 -04:00
Alexander Camuto
58ad46d74b
fix: sigmoid OOR issues ( #35 )
...
* OOR example
* An mnist with max pooling
* clean up scaling
* Update ezkl.rs
* cleanup further
* Update ezkl.rs
* trace out values
* bits and scale as cli params
* input ex
* relusig example
* add default pair for lookups
* fix sigmoid scaling
* Update README.md
* better logging of unknown funcs
* fix sigmoid tests
* defaults for bits and scale
* cli flag for k and default k
Co-authored-by: Jason Morton <jason.morton@gmail.com >
2022-10-07 13:49:42 -04:00
Alexander Camuto
3ef3c468c8
fix: conv rescaling and unknown values for fixed tensors ( #33 )
...
Co-authored-by: jason <jason.morton@gmail.com >
2022-10-05 23:21:14 +01:00
Jason Morton
29df95f19c
feat: load a model entirely from an onnx file and build circuit at runtime ( #25 )
...
Co-authored-by: Alexander Camuto <alexander.camuto@st-hughs.ox.ac.uk >
2022-10-04 14:49:03 +01:00
Alexander Camuto
78878e3f69
refactor: eltwise as layer ( #29 )
2022-09-30 18:01:17 +01:00
Alexander Camuto
815a5cd324
chore: documentation ( #28 )
...
* val and var tensor docstrings
* doc strings for tensor ops
* layer doc strings
* consistent interfaces for io and eltwise
* io docstrings
* cleanup
* onnx docs
* rm IOConfig and replace with VarTensor methods
* phrasing
2022-09-28 13:46:02 -04:00
Alexander Camuto
7a51bee68c
feat: benchmarks for relu, affine, cnvrl ( #24 )
...
* feat: benchmarks for varying size affine and conv layers
* relu benchmark
* cleanup
* fix mlp ex
2022-09-25 17:43:06 -04:00
Alexander Camuto
50a95f5cf2
refactor: fully dynamic conv layers ( #22 )
...
* chore: cleaner handling of 1D inputs/kernels
* fully dynamic conv
TODO: fix 1D kernel assignments
* support for 1D io
* mc enum test
* rmv visibility on helper functions
2022-09-23 09:26:17 -04:00
Alexander Camuto
76ec430dc5
chore: cleanup layers and update readme ( #20 )
...
* cleanup eltwise and conv
* phrasing
2022-09-20 14:57:49 -04:00
Alexander Camuto
55df4c63e6
feat: eltwise modular inputs and onnx model loading ( #14 )
...
Co-authored-by: jason <jason.morton@gmail.com >
2022-09-19 21:28:51 +01:00
Alexander Camuto
3a57438ee0
refactor: modular layers ( #13 )
...
* cleanup naming
* Update input.rs
* io layer
* fixed and advice layers
* working conv example
* cleanup
* fix value / prev assigned bug
* cleanup
* layer trait
* layer impl for affine and cnvrl
* reintegrate biases
* kernel length checks
* fix conv example shapes
2022-09-16 13:18:42 -04:00
Alexander Camuto
c62fd0a892
feat: tensors for multi-dimensional data ( #12 )
...
* feat: tensor types of arbitrary dims
* remove unused affine1d funcs
* restore forward
* tensor eltwise
* fix: lookup table mismatch
* iterator traits on tensor
* conv with tensors
* replace assign helper with enum map
* remove conv generics
* Delete all.txt
* fix slicing for tensors with singular dims
* allow single dim kernels
* rm matmul
* Update affine1d.rs
* proper 2d indexing for images and kernels
* indexing fixed
* Update mod.rs
* conv2d mnist example using tensors
* Update main.rs
* Update kernel.rs
* fix affine 1D
* infer affine out dim from weights
* cleanup
* conv2d new public input
* larger out dim
2022-09-07 15:52:47 -04:00
Alexander Camuto
63a0f40588
refactor: examples, subpackages ( #10 )
...
* refactor: move examples into own folder
* Delete mlp.rs
* Update lib.rs
* preliminary subfolders
* Update eltwise.rs
* Delete mlprescale.rs
* Update README.md
* Update README.md
Co-authored-by: Jason Morton <jason.morton@gmail.com >
2022-09-02 19:29:02 -04:00
