Xilinx/onnx-mlir
Feb 2025 – Apr 2025
2 Months active
Languages Used
C++MLIR
Technical Skills
Compiler DevelopmentGraph TransformationMLIRONNX RuntimeC++ DevelopmentONNX
Abhishek Bisht
PROFILE
Abhishek Bisht
Contributed to the Xilinx/onnx-mlir repository by developing two compiler features focused on graph transformation and shape inference. Implemented RMS Layer Normalization recomposition by detecting Pow operations with exponent -0.5, enabling more flexible and optimized graph transformations within the ONNX-MLIR backend. Additionally, delivered ONNX STFT shape inference support, allowing accurate propagation of output tensor dimensions based on input signals, frame lengths, and window sizes. Both features were integrated using C++ and MLIR, aligning with the ONNX dialect infrastructure. This work improved static analysis and optimization opportunities in the compiler pipeline, supporting more robust and efficient model transformations.
Overall Statistics
Feature vs Bugs
100%Features
Repository Contributions
2Total
Bugs
0
Commits
2
Features
2
Lines of code
422
Activity Months2
Your Network
64 peopleShared Repositories
64
Work History
April 2025
1 Commits • 1 Features
Apr 1, 2025April 2025 monthly summary focusing on key accomplishments and business value. The major feature delivered this month is the ONNX STFT shape inference support integrated into the Xilinx/onnx-mlir project. This work enables accurate propagation of output tensor dimensions for the STFT operation based on input signals, frame lengths, and window sizes, reducing runtime errors and improving optimization opportunities across the MLIR-based toolchain. The STFT feature was added to the ONNX dialect's build and the shape inference logic was implemented in C++, aligning with existing dialect infrastructure and build processes.
1 Commits • 1 Features
Apr 1, 2025April 2025 monthly summary focusing on key accomplishments and business value. The major feature delivered this month is the ONNX STFT shape inference support integrated into the Xilinx/onnx-mlir project. This work enables accurate propagation of output tensor dimensions for the STFT operation based on input signals, frame lengths, and window sizes, reducing runtime errors and improving optimization opportunities across the MLIR-based toolchain. The STFT feature was added to the ONNX dialect's build and the shape inference logic was implemented in C++, aligning with existing dialect infrastructure and build processes.
April 2025
February 2025
1 Commits • 1 Features
Feb 1, 2025February 2025 monthly summary for Xilinx/onnx-mlir: Delivered a feature to recomposition RMS Layer Normalization by recognizing a pattern containing Pow with exponent -0.5 to express reciprocal square root, enabling flexible and potentially optimized graph transformations. This work lays groundwork for streamlined RMSNormalisation handling and future performance optimizations across models using layer normalization.
February 2025
1 Commits • 1 Features
Feb 1, 2025February 2025 monthly summary for Xilinx/onnx-mlir: Delivered a feature to recomposition RMS Layer Normalization by recognizing a pattern containing Pow with exponent -0.5 to express reciprocal square root, enabling flexible and potentially optimized graph transformations. This work lays groundwork for streamlined RMSNormalisation handling and future performance optimizations across models using layer normalization.
Activity
Loading activity data...
Quality Metrics
Correctness95.0%
Maintainability90.0%
Architecture95.0%
Performance90.0%
AI Usage30.0%
Skills & Technologies
Programming Languages
C++MLIR
Technical Skills
C++ DevelopmentCompiler DevelopmentGraph TransformationMLIRONNXONNX RuntimeShape Inference
Repositories Contributed To
Overview of all repositories you've contributed to across your timeline