Oguzhan Gencoglu
Oguzhan Gencoglu
- [mlc](https://mlc.ai/mlc-llm/docs/) - [CTranslate2](https://github.com/OpenNMT/CTranslate2) - [FlexFlow](https://github.com/flexflow/FlexFlow) - [LMDeploy](https://github.com/InternLM/lmdeploy) - [skypilot](https://skypilot.readthedocs.io/en/latest/) - [Intel extension](https://github.com/intel/intel-extension-for-transformers)
Is your work a drop-in replacement for Adam or SGD or do I misunderstand it? If so, where is the source code? Do you have a pip library on pypi?
No releases or updates for years, unanswered issues, no support for new Python verisons. If the project is abandoned, it is better to say it explicitly on top of Readme.
Can benchmarks be updated for all the latest versions of libraries?
Add trankit performance for comparison: https://trankit.readthedocs.io/en/latest/ssplit.html
## Feature description OpenCV resize is way too slow for large images (e.g. 10k x 8k). `pyvips` is optimized for that. Ability to choose the resize backend would be great.
**Component** PyDIP - 3.5.1 (latest) **Describe the bug** I have a long python script that performs various computations with numpy, pandas, scipy etc. At one place, I have replaced my...
Would be useful to have an understanding of run-time performance of the most common morphological operations such as - erosion - dilation - thinning - closing - opening - skeletonization...
Do you support ONNX export (similar to `skl2onnx`)? Is there any example docs for that?
Do gradients and Hessians accept `weight` for custom regression loss? For example your docs do not mention that: ``` def grad_msle( labels: npty.NDArray[np.float32], predictions: npty.NDArray[np.float32] ) -> npty.NDArray[np.float32]: gradient =...