GitHub / modern-fortran/neural-fortran / commits
| SHA | Message | Author | Date | Stats |
|---|---|---|---|---|
| 00acae24 | fix #233 - make factories impure (#234) |
Damian Rouson <r****n@l****v>
Committed by: GitHub <n****y@g****m> |
about 1 month ago | |
| 3049393b | merge Merge pull request #232 from jvdp1/get_activation |
Jeremie Vandenplas <j****s@g****m>
Committed by: GitHub <n****y@g****m> |
about 1 month ago | |
| 37cd2d90 | Removed non-required use | Jeremie Vandenplas <j****s@g****m> | about 1 month ago | |
| f02a6fd8 | nf_dense_layer: removed unused set_params | Jeremie Vandenplas <j****s@g****m> | about 1 month ago | |
| 6ed81fd1 | Addition of get_optimizer_by_name and of get_name for optimizer DT | Jeremie Vandenplas <j****s@g****m> | about 1 month ago | |
| 5110fc6d | Set get_activation by name to pure | Jeremie Vandenplas <j****s@g****m> | about 1 month ago | |
| 4bfa3078 | Removed unused dt | Jeremie Vandenplas <j****s@g****m> | about 1 month ago | |
| 89a255be |
Fixed compilation errors for Intel compilers (#229)
Co-authored-by: Vandenplas, Jeremie <j****s@w****l> |
Jeremie Vandenplas <j****s@g****m>
Committed by: GitHub <n****y@g****m> |
about 2 months ago | |
| 3658b8a8 | merge Merge pull request #227 |
Riccardo Orsi <1****1@u****m>
Committed by: GitHub <n****y@g****m> |
2 months ago | |
| 226b0306 | Removed set_params as well | Riccardo Orsi <r****i@i****t> | 3 months ago | |
| d36edc78 | First change; take a look | Riccardo Orsi <r****i@i****t> | 3 months ago | |
| b2073fa5 | Rename locally_connected1d -> locally_connected2d (#224) |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
3 months ago | |
| 1c968ceb | Apply optimizer to model weights without data copy (#222) |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
3 months ago | |
| 402b84ae |
Generic conv & maxpool (#220)
Co-authored-by: Jeremie Vandenplas <j****s@g****m>, Jeremie Vandenplas <j****s@g****m>, Jeremie Vandenplas <j****s@g****m>, Jeremie Vandenplas <j****s@g****m>, Jeremie Vandenplas <j****s@g****m> |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
6 months ago | |
| 2ed7b6a8 | Multihead Attention Fixes (#209) |
Michael Voronov <m****o@l****u>
Committed by: GitHub <n****y@g****m> |
7 months ago | |
| 1db02588 | Generic reshape (#213) |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
8 months ago | |
| 5cf14ebd |
Locally connected layer (#201)
Co-authored-by: Jeremie Vandenplas <j****s@g****m>, Jeremie Vandenplas <j****s@g****m>, milancurcic <c****o@g****m> |
Riccardo Orsi <1****7@u****m>
Committed by: GitHub <n****y@g****m> |
8 months ago | |
| e628d1ed |
Embedding Layer (#205)
Co-authored-by: milancurcic <c****o@g****m> |
Michael Voronov <m****o@l****u>
Committed by: GitHub <n****y@g****m> |
8 months ago | |
| e68e6c2d |
Layernorm (#203)
Co-authored-by: milancurcic <c****o@g****m> |
Michael Voronov <m****o@l****u>
Committed by: GitHub <n****y@g****m> |
8 months ago | |
| ed8b3404 |
Multihead attention (#199)
Co-authored-by: Jeremie Vandenplas <j****s@g****m>, Jeremie Vandenplas <j****s@g****m>, Jeremie Vandenplas <j****s@g****m>, Jeremie Vandenplas <j****s@g****m>, milancurcic <c****o@g****m>, Jeremie Vandenplas <j****s@g****m> |
Michael Voronov <m****o@l****u>
Committed by: GitHub <n****y@g****m> |
8 months ago | |
| 039638d9 |
Dropout layer (#194)
Co-authored-by: Ricardo Orsi <****@r****7>, Vandenplas, Jeremie <j****s@w****l> |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
8 months ago | |
| c316ee1e |
Linear2d layer (#197)
Co-authored-by: milancurcic <c****o@g****m> |
Michael Voronov <m****o@l****u>
Committed by: GitHub <n****y@g****m> |
9 months ago | |
| 4ad75bc3 |
Generic flatten (2d and 3d) (#202)
Co-authored-by: Mikhail Voronov <m****o@l****u> |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
9 months ago | |
| a28a9be2 |
Add Input2d layer (#198)
Co-authored-by: Mikhail Voronov <m****o@l****u> |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
9 months ago | |
| d5164378 |
Update for cmake use of neural-fortran (#192)
Co-authored-by: Milan Curcic <c****o@g****m> |
Matt Thompson <f****n@g****m>
Committed by: GitHub <n****y@g****m> |
about 1 year ago | |
| 9a401171 | Standard conformance fixes (#190) |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
about 1 year ago | |
| 38722021 | Remove Keras HDF5 (#188) |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
about 1 year ago | |
| 118f795d |
Intrinsic `pack` replaced by pointers in `get_params` and `get_gradients` (#183)
Co-authored-by: Vandenplas, Jeremie <j****s@w****l>, milancurcic <c****o@g****m> |
Jeremie Vandenplas <j****s@g****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| a843c83d |
Remove some unused variables and a broadcast (#180)
Co-authored-by: Vandenplas, Jeremie <j****s@w****l> |
Jeremie Vandenplas <j****s@g****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| e82d5652 |
Proposition of API for the method `network % evaluate` (#182)
Co-authored-by: Vandenplas, Jeremie <j****s@w****l>, milancurcic <c****o@g****m> |
Jeremie Vandenplas <j****s@g****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 6dfaed00 |
Replacement of a matmul + use of merge (#181)
Co-authored-by: Vandenplas, Jeremie <j****s@w****l>, milancurcic <c****o@g****m> |
Jeremie Vandenplas <j****s@g****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| c3924b51 | Addidtion of Findneural-fortran.cmake (#178) |
Jeremie Vandenplas <j****s@g****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| f7b6006f |
Addition of the Loss derived type and of the MSE loss function (#175)
Co-authored-by: Vandenplas, Jeremie <j****s@w****l>, milancurcic <c****o@g****m> |
Jeremie Vandenplas <j****s@g****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| cf471147 | Add instructions using Conda (#172) |
Ondřej Čertík <o****j@c****s>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| ca219cd3 | More conv2d tests (#174) |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| b1b2cacb |
safegaurd Box-Muller normal random number generation against u=0.0 (#158)
Co-authored-by: milancurcic <c****o@g****m> |
dacarnazzola <1****a@u****m>
Committed by: GitHub <n****y@g****m> |
about 2 years ago | |
| b119194a |
Adagrad Optimizer Implementation (#154)
Co-authored-by: milancurcic <c****o@g****m> |
Saurabh Suresh Powar <6****5@u****m>
Committed by: GitHub <n****y@g****m> |
about 2 years ago | |
| 6adc1c25 |
Added Adam optimizer implementation (#150)
Co-authored-by: milancurcic <c****o@g****m> |
Saurabh Suresh Powar <6****5@u****m>
Committed by: GitHub <n****y@g****m> |
over 2 years ago | |
| e9bfbd60 |
Added Momentum and Nesterov modifications (#148)
Co-authored-by: milancurcic <c****o@g****m> |
Saurabh Suresh Powar <6****5@u****m>
Committed by: GitHub <n****y@g****m> |
over 2 years ago | |
| 6bbc28d1 | Connect `flatten`, `conv2d`, and `maxpool2d` layers in backward pass (#142) |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
over 2 years ago | |
| 31fc0617 | SGD optimizer stub (#139) |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
over 2 years ago | |
| 44833c20 |
Added RMSProp Optimizer subroutine (#144)
Co-authored-by: milancurcic <c****o@g****m> |
Saurabh Suresh Powar <6****5@u****m>
Committed by: GitHub <n****y@g****m> |
over 2 years ago | |
| c3c048ac | add CELU activation function (#143) |
Pablo del Mazo Sevillano <4****o@u****m>
Committed by: GitHub <n****y@g****m> |
over 2 years ago | |
| f9ff658a | Add example program on how to access internal layer parameters (#140) |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
over 2 years ago | |
| e34671e0 | Add missing update for conv2d_layer (#141) |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
over 2 years ago | |
| f6c08546 | Update contributors and acknowledgment (#138) |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
over 2 years ago | |
| adaf9bb2 |
GSoC Optimizers: Example program to fit a quadratic function (#134)
Co-authored-by: milancurcic <c****o@g****m> |
Saurabh Suresh Powar <6****5@u****m>
Committed by: GitHub <n****y@g****m> |
over 2 years ago | |
| 82931182 | Contributing guide (#109) |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
over 2 years ago | |
| d052bce6 |
Class-based activation functions (#126)
Co-authored-by: milancurcic <c****o@g****m> |
Goyman Gordey <4****n@u****m>
Committed by: GitHub <n****y@g****m> |
over 2 years ago | |
| f328c8df |
Added Leaky ReLU activation function (1D & 3D) (#123)
Co-authored-by: Milan Curcic <c****o@g****m> |
Saurabh Suresh Powar <6****5@u****m>
Committed by: GitHub <n****y@g****m> |
over 2 years ago | |
| 7039276b | Update contributors (#120) |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
almost 3 years ago | |
| 0b916ee8 |
Add linear activation function (#119)
Co-authored-by: Richard Weed <r****d@c****u> |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
almost 3 years ago | |
| e51a7465 | Insert a flatten layer if a dense layer follows a layer (#118) |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
almost 3 years ago | |
| edd3f70a | Introduce optimizer_base_type in support of different optimizers (#116) |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
almost 3 years ago | |
| dce3b0a6 | Small workaround to make Intel compiler happy (#117) |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
almost 3 years ago | |
| 3588d075 | Add missing edit descriptor (#113) |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
almost 3 years ago | |
| 8fb2ef38 |
Get set network parameters (#111)
Co-authored-by: Christopher Zapart <c****t@n****p>, Christopher Zapart <c****t@n****p>, Christopher Zapart <c****t@n****p> |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
almost 3 years ago | |
| 1b0646a2 |
Fix CMake build; Co-authored-by: jvo203 <chris.zapart@nao.ac.jp> (#110)
Co-authored-by: jvo203 <c****t@n****p> |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
almost 3 years ago | |
| 9bbd70f2 | CNN backward pass (#99) |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
about 3 years ago | |
| e9af5b43 | Implement the reshape layer (#97) |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
about 3 years ago | |
| 956c28a8 | Bump version in fpm.toml (#96) |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
about 3 years ago | |
| 79b820ec | Fix accidental TOC reorder from a previous PR (#94) |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
about 3 years ago | |
| 13617e89 | Update README (#93) |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
about 3 years ago | |
| 51aad71f | merge Merge pull request #92 from milancurcic/rename-output-to-predict |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
about 3 years ago | |
| 78f762f1 | Rename output -> predict for consistency with Keras | milancurcic <c****o@g****m> | about 3 years ago | |
| 20ec3499 | merge Merge pull request #90 from milancurcic/batch-inference |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
over 3 years ago | |
| b7108722 | Specific procedures for batch output | milancurcic <c****o@g****m> | over 3 years ago | |
| 191aec96 | Add specific network methods for batch forward and output | milancurcic <c****o@g****m> | over 3 years ago | |
| 75666fe5 | merge Merge pull request #87 from milancurcic/ci |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
over 3 years ago | |
| ca9c64a7 | More concise job names | Milan Curcic <c****o@g****m> | over 3 years ago | |
| f51c78ef | Try alternative syntax | Milan Curcic <c****o@g****m> | over 3 years ago | |
| 6e65be36 | Debug and release profiles | Milan Curcic <c****o@g****m> | over 3 years ago | |
| 447b5ce7 | Specify profile | Milan Curcic <c****o@g****m> | over 3 years ago | |
| d7e833ab | First stab at build and test on Ubuntu+fpm in CI | Milan Curcic <c****o@g****m> | over 3 years ago | |
| a878d038 | merge Merge pull request #85 from milancurcic/keras-reader-cnn |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
over 3 years ago | |
| 0577f821 | Update list of dependencies | milancurcic <c****o@g****m> | over 3 years ago | |
| c0baf373 | Update the project description (not so micro anymore) | milancurcic <c****o@g****m> | over 3 years ago | |
| eb992cb4 | Bump version to 0.6.0 | milancurcic <c****o@g****m> | over 3 years ago | |
| 0ada76a8 | Update the README feature and examples list | milancurcic <c****o@g****m> | over 3 years ago | |
| 148892bd | Add CNN from Keras example; rename the dense from Keras example | milancurcic <c****o@g****m> | over 3 years ago | |
| 0941ebbb | Add CMake rules for functional-fortran | milancurcic <c****o@g****m> | over 3 years ago | |
| 3a5cc190 | Remove unused import | milancurcic <c****o@g****m> | over 3 years ago | |
| 925e3064 | Allow as 3-d output of networks; assign activation function to conv2d layers ... | milancurcic <c****o@g****m> | over 3 years ago | |
| 5592b9a4 | CNN from Keras test is passing | milancurcic <c****o@g****m> | over 3 years ago | |
| 2ff343ec | Don't transpose 4-d kernel; already stored correctly | milancurcic <c****o@g****m> | over 3 years ago | |
| 0268b266 | Use reshape to transpose n-dim HDF5 arrays | milancurcic <c****o@g****m> | over 3 years ago | |
| eb70debb | Read kernel and bias data for Conv2D layer from Keras; tranpose values in the... | milancurcic <c****o@g****m> | over 3 years ago | |
| 87ebc11a | Remove redundant if-block | milancurcic <c****o@g****m> | over 3 years ago | |
| ff2321de | Enable Conv2D, Flatten, and MaxPooling2D in the network constructor from Keras | milancurcic <c****o@g****m> | over 3 years ago | |
| d943b0c1 | Start the CNN from Keras test suite | Milan Curcic <c****o@g****m> | over 3 years ago | |
| df3d8e2b | Handle C->Fortran order with Keras layer dims | Milan Curcic <c****o@g****m> | over 3 years ago | |
| 44cd48d0 | Add MaxPooling2D and Flatten layers to the Keras reader | Milan Curcic <c****o@g****m> | over 3 years ago | |
| 878e973d | Rename keras_layer attribute for consistency with Keras notation | Milan Curcic <c****o@g****m> | over 3 years ago | |
| 336eabba | Generalize input shape in Keras reader; allow reading Conv2D metadata | milancurcic <c****o@g****m> | over 3 years ago | |
| 915cec52 | Add URL to Keras CNN MNIST model; make URL constants more concise | milancurcic <c****o@g****m> | over 3 years ago | |
| 94cc86be | merge Merge pull request #84 from modern-fortran/update-contributors-list |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
over 3 years ago | |
| a971fd49 | Update contributors list |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
over 3 years ago | |
| 0cbf797e | merge Merge pull request #83 from jacobwilliams/82-memory-leak-hotfix |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
over 3 years ago | |
| b481a5bf | fixed a memory leak when reading the JSON | Jacob Williams <j****s@u****m> | over 3 years ago | |
| 19492dd0 | merge Merge pull request #81 from milancurcic/add-citation-cff |
Milan Curcic <c****o@g****m>
Committed by: GitHub <n****y@g****m> |
over 3 years ago |