I'm a hobby programmer. 
Intel Software Innovator member list
PINTO0309
Follow
Pinned
5,302 contributions in the last year
Activity overview
Contributed to
PINTO0309/onnx2tf,
PINTO0309/PINTO_model_zoo,
PINTO0309/simple-onnx-processing-tools
and 132 other
repositories
Contribution activity
March 2023
Created 122 commits in 6 repositories
Created a pull request in PINTO0309/onnx2tf that received 1 comment
1. Temporarily disabled accuracy check function added as a backend function. 2. Backward to v1.13.1 because I noticed a bug in sigmoid behavior in onnxruntime v1.14.0
1. Content and background Temporarily disabled accuracy check function added as a backend function. #184 Significantly reduced RAM consumption du…
+350
−318
•
1
comment
Opened 21 other pull requests in 2 repositories
PINTO0309/onnx2tf
1
open
19
merged
-
Fixed a problem with the
shape_unmatched_special_avoidance_workaroundjudgment condition to improve the Transformer conversion success rate. -
Improved INT8 quantization to be performed based on
saved_modelsignature information - [Experimental] Deterring shape corruption due to broadcast. Transformer conversion stability improvement. (3D and below only)
-
TensorFlow's tf.math.cumsum does not support boolean, so temporarily cast to INT32 for calculation.
CumSum -
Fixed out-of-range reference errors for begin and end when executing
Slicefor padding removal afterConvTransposegeneratio -
Improved stability of shape estimation when onnxsim is not yet executed and fixed a bug in parameter replacement process.
ConvTranspose -
Improved stability of type estimation when onnxsim is not yet executed.
Pow - 1.Improved stability of shape estimation when onnxsim is not yet executed. 2.Improved processing order of parameter replacement process
-
Support for
ScaleAndTranslate -
Support for
preserve_aspect_ratioandantialiasResize.opset >= 18 -
Improved stability of model transformation when onnxsim is not running.
ReduceMax,ReduceMean,ReduceMin,ReduceProd - Correction of accuracy check logic breakage due to operation merging process
- Merge two consecutive identical OPs into one
-
-coion / --copy_onnx_input_output_names_to_tfliteoutput tflite with options so that the signature is embedded in the tflite by default. - Prevent saved_model (.pb) bloat
-
Suppress automatic
FlexTrasposeextrapolation behavior bySoftmaxin TensorFlow -
Fixed
Softmaxaccuracy check bug and reducedPowcalculation error - Improved error judgment regarding calibration data for INT8 quantization
-
Disabling splits that do not make sense and preventing the generation of
FlexSplitVwere added. - Corrected a mistake in the decision statement
PINTO0309/sit4onnx
1
merged
Created an issue in PINTO0309/onnx2tf that received 1 comment
[TODO] Implement the Resize operation above opset=18
Issue Type Others onnx2tf version number 1.7.21 onnx version number 1.13.1 tensorflow version number 1.12.0rc1 Download URL for ONNX N/A Parameter …
1
comment
Opened 5 other issues in 2 repositories
PINTO0309/onnx2tf
2
open
2
closed
microsoft/onnxruntime
1
open
102
contributions
in private repositories
Mar 1 – Mar 8