sdpython/onnxcustom issues and pull requests

#91 - Support for FN, UZ variant of float 8 types

Pull Request - State: closed - Opened by xadupre almost 2 years ago

#90 - Improves float 8 functions and documentation

Pull Request - State: closed - Opened by xadupre almost 2 years ago
Labels: documentation

#89 - Supports slice operator in onnx_split

Pull Request - State: open - Opened by sdpython almost 2 years ago
Labels: enhancement

#88 - Fixes backward propagation for shape results

Pull Request - State: closed - Opened by sdpython about 2 years ago
Labels: bug

#87 - Implements conversion with E5M2

Pull Request - State: closed - Opened by sdpython about 2 years ago
Labels: enhancement

#86 - Implementations for float 8 conversion

Pull Request - State: closed - Opened by xadupre about 2 years ago
Labels: enhancement

#85 - Adds one example, update unit tests

Pull Request - State: closed - Opened by xadupre about 2 years ago

#84 - Adds a unit test about QAT

Pull Request - State: closed - Opened by xadupre about 2 years ago
Labels: ci

#83 - Adds example on quantization

Pull Request - State: closed - Opened by xadupre about 2 years ago
Labels: documentation

#82 - Blog post

Pull Request - State: closed - Opened by xadupre about 2 years ago
Labels: documentation

#81 - Documentation, check ai.onnx.ml opset 3 on lightgbm

Pull Request - State: closed - Opened by xadupre about 2 years ago
Labels: documentation

#80 - Improves the splitting shape nodes in split_onnx

Pull Request - State: closed - Opened by xadupre about 2 years ago
Labels: bug

#79 - Update Windows CI on python 3.10, fix one example

Pull Request - State: closed - Opened by xadupre about 2 years ago
Labels: documentation, ci

#78 - Uses new method run_with_ortvaluevector

Pull Request - State: closed - Opened by xadupre about 2 years ago
Labels: documentation

#77 - Adds an example to benchmark a kind of eager mode

Pull Request - State: closed - Opened by xadupre about 2 years ago
Labels: documentation

#76 - MatMul reshape Gradient computation

Issue - State: closed - Opened by cjm-cai about 2 years ago - 1 comment

#75 - Supports small inputs in onnx_split

Pull Request - State: closed - Opened by xadupre about 2 years ago
Labels: bug

#74 - Supports optional inputs in split_onnx

Pull Request - State: closed - Opened by xadupre about 2 years ago
Labels: bug

#73 - Update CI

Pull Request - State: closed - Opened by xadupre about 2 years ago
Labels: ci

#72 - Support for custom cutting points

Pull Request - State: closed - Opened by xadupre about 2 years ago
Labels: enhancement

#71 - Splits onnx models

Pull Request - State: closed - Opened by sdpython about 2 years ago
Labels: enhancement

#70 - Modifies one example about parallel execution

Pull Request - State: closed - Opened by xadupre about 2 years ago
Labels: documentation

#69 - Updates OrtGradientForwardBackward to support the latest onnxruntime

Pull Request - State: closed - Opened by xadupre about 2 years ago
Labels: bug

#68 - Adds an example to check parallel execution

Pull Request - State: closed - Opened by xadupre over 2 years ago
Labels: documentation

#67 - Add one more example about TfidfVectorizer

Pull Request - State: closed - Opened by xadupre over 2 years ago
Labels: documentation

#66 - Uses f strings

Pull Request - State: closed - Opened by xadupre over 2 years ago
Labels: code

#65 - Adds one example to benchmark onnx serialization functions

Pull Request - State: closed - Opened by xadupre over 2 years ago
Labels: documentation

#64 - Fixes bug introduced by recent updates

Pull Request - State: closed - Opened by xadupre over 2 years ago
Labels: bug

#63 - Add GraphProto to the documentation

Issue - State: closed - Opened by xadupre almost 3 years ago
Labels: documentation

#62 - Adds a notebook about convolution

Pull Request - State: closed - Opened by xadupre almost 3 years ago
Labels: documentation

#61 - Update CI to use latest onnxruntime-training

Pull Request - State: closed - Opened by xadupre almost 3 years ago
Labels: ci

#60 - Extends notebook coverage

Pull Request - State: closed - Opened by xadupre almost 3 years ago
Labels: ci

#59 - Removes unnecessary exceptions

Pull Request - State: closed - Opened by sdpython almost 3 years ago
Labels: bug

#58 - Uses new API to retrieve gradient for a model

Pull Request - State: closed - Opened by sdpython almost 3 years ago
Labels: enhancement

#57 - Fix opset for ai.onnx.ml in examples after updating to onnx==1.11

Pull Request - State: closed - Opened by xadupre almost 3 years ago
Labels: bug

#56 - Renames Z into Y_grad in loss functions returning grad

Pull Request - State: closed - Opened by xadupre almost 3 years ago
Labels: refactoring

#55 - Removes unncessary nodes in onnx_derivative

Pull Request - State: closed - Opened by xadupre almost 3 years ago
Labels: bug

#54 - Implements a function which returns the gradient

Pull Request - State: closed - Opened by sdpython about 3 years ago
Labels: enhancement

#53 - Implements get_trained_onnx to retrieve the trained model

Pull Request - State: closed - Opened by sdpython about 3 years ago
Labels: enhancement

#52 - [WONT] Allows DataLoader to fit, predict

Issue - State: closed - Opened by sdpython about 3 years ago
Labels: enhancement

#51 - Implements scoring functions

Pull Request - State: closed - Opened by sdpython about 3 years ago
Labels: enhancement

#50 - [WONT] Cch2

Pull Request - State: closed - Opened by sdpython about 3 years ago

#49 - Check nan values during training

Pull Request - State: closed - Opened by sdpython about 3 years ago
Labels: bug

#48 - Checks gradients are identical to scikit-learn for neural networks

Pull Request - State: closed - Opened by sdpython about 3 years ago
Labels: enhancement

#47 - Compare gradient values with scikit-learn

Issue - State: closed - Opened by sdpython about 3 years ago
Labels: enhancement

#46 - Deal with nan values during training

Issue - State: open - Opened by sdpython about 3 years ago
Labels: bug

#45 - Adds an example about classification

Pull Request - State: closed - Opened by sdpython about 3 years ago
Labels: documentation

#44 - Uses bind_ortvalue_input instead of bind_input to be faster

Pull Request - State: closed - Opened by sdpython about 3 years ago
Labels: performance

#43 - Fixes the training of a binary classifier with weights

Pull Request - State: closed - Opened by sdpython about 3 years ago
Labels: bug

#42 - Implements binary log loss for first API of orttraining

Pull Request - State: closed - Opened by xadupre about 3 years ago
Labels: enhancement

#41 - Refactors to improve profiling analysis (fix penalty, export onnx graphs)

Pull Request - State: closed - Opened by xadupre about 3 years ago
Labels: refactoring

#40 - Improves training fwbw with caching.

Pull Request - State: closed - Opened by sdpython about 3 years ago
Labels: performance

#39 - Introduce large files with ONNX

Issue - State: open - Opened by sdpython about 3 years ago
Labels: documentation

#38 - Use C class InferenceSession and TrainingSession

Issue - State: open - Opened by sdpython about 3 years ago
Labels: performance

#37 - Improves performance of caching

Pull Request - State: closed - Opened by sdpython about 3 years ago
Labels: performance

#36 - Makes sure all plt.show() have been disabled in examples

Pull Request - State: closed - Opened by xadupre about 3 years ago
Labels: documentation

#35 - Reduces the number of calls to bind_ortvalue

Pull Request - State: closed - Opened by xadupre about 3 years ago
Labels: performance

#34 - Implements L1, L2 losses for partial training

Pull Request - State: closed - Opened by sdpython about 3 years ago
Labels: enhancement

#33 - Implements more loss functions

Pull Request - State: closed - Opened by sdpython about 3 years ago
Labels: enhancement

#32 - Implements classificiation loss

Issue - State: closed - Opened by sdpython about 3 years ago
Labels: enhancement

#31 - Implements penalty when running the gradient

Issue - State: closed - Opened by sdpython about 3 years ago
Labels: enhancement

#30 - Implements different learning rate strategies

Pull Request - State: closed - Opened by sdpython about 3 years ago
Labels: enhancement

#29 - Fixes #16, support weights while training

Pull Request - State: closed - Opened by sdpython about 3 years ago
Labels: enhancement

#28 - Replaces OrtValue by C_OrtValue everywhere

Pull Request - State: closed - Opened by sdpython about 3 years ago
Labels: refactoring

#27 - Uses C_OrtDevice everywhere

Pull Request - State: closed - Opened by sdpython about 3 years ago
Labels: refactoring

#26 - Implements learning rate from neural network

Issue - State: closed - Opened by xadupre about 3 years ago
Labels: enhancement

#25 - Implements optimizers with forward, backward functionalities

Pull Request - State: closed - Opened by sdpython about 3 years ago
Labels: enhancement

#24 - Be more consistent with OrtValue, OrtDevice, C and python versions

Issue - State: closed - Opened by sdpython about 3 years ago
Labels: refactoring

#23 - Extend documentation

Pull Request - State: closed - Opened by xadupre about 3 years ago
Labels: documentation

#22 - Move learning_rate logic in a separate class

Pull Request - State: closed - Opened by xadupre about 3 years ago
Labels: enhancement

#21 - Implements a mechanism that update training weights with SGDRegressor or MLPRegressor

Issue - State: closed - Opened by xadupre about 3 years ago
Labels: enhancement

#20 - Experiment markdown rendering

Pull Request - State: closed - Opened by xadupre about 3 years ago
Labels: documentation

#19 - Implements training with forward, backward

Pull Request - State: closed - Opened by sdpython about 3 years ago
Labels: enhancement

#18 - Adds classes to train an ONNX gradient with TrainingAgent

Pull Request - State: closed - Opened by sdpython about 3 years ago
Labels: enhancement

#17 - Minimize the number of data copy while training a model

Pull Request - State: closed - Opened by xadupre about 3 years ago
Labels: performance

#16 - Support weights when training a model

Issue - State: closed - Opened by xadupre about 3 years ago

#15 - Add example with TrainingAgent (error gradient outside)

Issue - State: closed - Opened by xadupre about 3 years ago

#14 - Optimize DataLoader to use iobinding (avoir copy)

Issue - State: closed - Opened by xadupre about 3 years ago - 1 comment

#13 - Adds function plot_onnxs

Pull Request - State: closed - Opened by xadupre about 3 years ago
Labels: documentation, enhancement

#12 - Add more examples

Pull Request - State: closed - Opened by xadupre about 3 years ago
Labels: documentation

#11 - Automates nvprof logs retrieval

Pull Request - State: closed - Opened by xadupre over 3 years ago
Labels: enhancement

#10 - Add parameter to evaluate the model on test data while training

Pull Request - State: closed - Opened by sdpython over 3 years ago
Labels: enhancement

#9 - Refactoring documentation

Pull Request - State: closed - Opened by sdpython over 3 years ago
Labels: documentation

#8 - Add example to look into neural network on GPU

Pull Request - State: closed - Opened by xadupre over 3 years ago
Labels: documentation

#7 - Refactoring

Pull Request - State: closed - Opened by sdpython over 3 years ago

#6 - Add examples with orttraining

Pull Request - State: closed - Opened by xadupre over 3 years ago

#5 - Fix examples, update documentation

Pull Request - State: closed - Opened by xadupre over 3 years ago
Labels: bug, documentation

#4 - Complex scenarios

Pull Request - State: closed - Opened by sdpython over 3 years ago
Labels: documentation

#3 - Replaces OnnxSubOperator by OnnxSubEstimator

Pull Request - State: closed - Opened by sdpython almost 4 years ago
Labels: deprecation

#2 - Add example of custom operator with onnxruntime

Issue - State: open - Opened by sdpython over 4 years ago
Labels: documentation

#1 - Add an example on black list, while list of operators.

Issue - State: closed - Opened by sdpython over 4 years ago
Labels: documentation

GitHub / sdpython/onnxcustom issues and pull requests