Fix NumPy 1.21 Cross-Compile in Buildroot 2019.02

₹12500-37500 INR

Sürüyor

İlan edilme: 3 ay önce

₹12500-37500 INR

Teslimde ödenir

We have an ARMV7 (32-bit) T113 All Winner development board and it supports Buildroot 2019.02 with Linux 5.4 along with Python 3.7 inbuilt support. Buildroot 2019.02 does not support python 3.9 version but we have already rebuilt with Python 3.9 on an Ubuntu 20.04 LTS workstation. The next hurdle is getting NumPy 1.21 to cross-compile cleanly, then bundling a small set of scientific-audio libraries into the same root fs. At the moment the build stalls with messages such as “toolchain C linking failed”, “Fortran required for compilation”, and the infamous “No module named 'numpy.core._multiarray_umath'”. These errors appear during the host-build detection phase as well as during target compilation, so the solution will likely involve patching Buildroot packages, tweaking [login to view URL], and perhaps supplying or disabling a cross-Fortran toolchain. Target Python stack that must end up in the image: • tflite-runtime >= [login to view URL] • numpy >= 1.21.0 • psutil >= 5.5.0 (only for internal diagnostics; will be stripped before release) • soundfile >= 0.13.1 • samplerate >= 0.1.0 What I need from you – Adjust or create Buildroot package definitions and associated [login to view URL] / *.mk files so NumPy 1.21 builds without manual intervention. – Resolve (or cleanly work around) the Fortran requirement while keeping the final image minimal. – Ensure the remaining Python wheels above cross-compile and import successfully on the target board. – Supply clear, repeatable build steps or a patch series I can apply to a fresh Buildroot 2019.02 tree. - Compiled [login to view URL] is what we require to build the final image. Acceptance criteria 1. Make defconfig && make completes on my side with no errors. 2. Running “python3 -c 'import numpy, soundfile, samplerate, psutil, tflite_runtime as tflite; print(numpy.__version__)'” on the target prints the expected versions. 3. Total root fs growth stays within reason; no unused build-time Fortran binaries left in the target image. 4. Compiled [login to view URL] is what we require to build the final image. If you have tackled NumPy or other heavy C/Fortran libraries in Buildroot before, this should feel familiar. Let me know what additional diagnostics you need, and I will provide full build logs or access to the current branch.

Software Architecture

Proje No: 40052543

Proje hakkında

13 teklifler

Uzaktan proje

Son aktiviteden bu yana geçen zaman 3 ay önce

Biraz para mı kazanmak istiyorsunuz?

E-posta adresi

Freelancer'da teklif vermenin faydaları

Bütçenizi ve zaman çerçevenizi belirleyin

Çalışmanız için ödeme alın

Teklifinizin ana hatlarını belirleyin

Kaydolmak ve işlere teklif vermek ücretsizdir

Seçilen:

@ahmadarajaa

₹0 INR 4 gün içinde

5,0

(4 değerlendirme)

3,4

13 freelancer bu proje için ortalama ₹26.050 INR teklif veriyor

@chraheelbhatti

I'm Rahil, a seasoned Machine Learning Engineer and Researcher with ample experience working on complex Linux projects like yours. Over the past 8 years, I have specialized in Computer Vision, NLP, Generative AI, and Cloud-based AI solutions. This makes me highly adept at managing projects involving heavy libraries such as NumPy and resolving compatibility issues that might arise during cross-compilation. Moreover, my skills extend beyond troubleshooting. I have led projects at Unilever Pakistan, State Bank of Pakistan and other important institutions in my country where I have had to ensure the smooth integration of complex AI systems into operational environments. This requires an attention to detail and the ability to optimize systems for minimal memory footprints - qualities that will be very helpful in dealing with Buildroot packages and keeping the final image size minimal as required in your project. Not only will I fix the issues with NumPy 1.21 Cross-Compile in Buildroot 2019.02 but also provide you with clear, repeatable build steps or a patch series you can easily apply to a fresh Buildroot 2019.02 tree to avoid any problems down the line. Beyond that, I'll make certain that the target Python stack imports seamlessly on the ARMV7 board and guarantee a final build that satisfies all your acceptance criteria. Take leverage of my strong command over Python, Linux & proven experience with NumPy, and award me your project;let's yield excellent results together.

₹37.500 INR 5 gün içinde

4,7

(24 değerlendirme)

5,5

@zaheermoe

Hi Amitkumarssharma, I understand the challenges you are facing with getting NumPy 1.21 to cross-compile cleanly on your ARMV7 development board running Buildroot 2019.02. With my experience in adjusting Buildroot package definitions, resolving Fortran requirements, and ensuring successful cross-compilation of Python wheels, I can help you achieve your goals efficiently. To overcome the hurdles you mentioned, I propose creating tailored Buildroot package definitions for NumPy 1.21, addressing the Fortran requirement effectively, and providing clear build steps or patches for a seamless integration. Let's collaborate to ensure a flawless build process and meet all acceptance criteria. Looking forward to working together on this project. My availability to start is immediate. Best, Zaheer Mahomed

₹23.650 INR 30 gün içinde

0,0

(0 değerlendirme)

0,0

@Thabs9

I'm excited to collaborate on fixing NumPy 1.21 cross-compile in Buildroot 2019.02 for your ARMV7 T113 All Winner board. With expertise in Python, Linux, and embedded systems, I'll adjust Buildroot packages to automate NumPy 1.21 building and resolve any Fortran issues. Your desired Python stack with tflite-runtime, numpy, psutil, soundfile, and samplerate will seamlessly cross-compile. I'll ensure error-free builds, reflect the expected versions accurately, and maintain a minimal final image. If needed, I can provide detailed build data. My past experience with these libraries in Buildroot makes me a dependable choice for this project. Eagerly anticipating our potential synergy on this task. Regards, Thabo

₹18.708 INR 3 gün içinde