New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Release 2024-01 follow-up #1274
Comments
|
Audit trail:
b1 (2023-12-17: numba-rc and pre-torch on Python-3.12)
b2 (2024-02-2x: ortools, still numba-rc and pre-torch on Python-3.12)
b3 (2024-03-0x )
rc (2024-03-1x )
|
|
WinPython 2024-01 b1 (2023-12-17) Changes from WinPython 2023-04:
Areas of particular interest for testers:
next build:
binaries at:
WinPython 3.11.7.0The following packages are included in WinPython-64bit v3.11.7.0 b1. Tools
Python packages
WinPython 3.12.1.0The following packages are included in WinPython-64bit v3.12.1.0 b1. Tools
Python packages
|
|
Playing with https://webllm.mlc.ai/ (so webgpu on a weak intel laptop of 4 years ago)
Both ugly, and yet, 1 year ago we had only 4 words per second on Cloud-only. |
|
Is |
Numba roadmap is every 3 month, so Numba-0.59 final can only be 3 month late, so shall be in next cycle. |
|
Prelude: python-3.13.0a2 (2023-12-31) Python-3.13.0a2 is not yet including jit, interpreters official package, free-threads... Areas of particular interest for testers:
Binaries at:
Next builds:
WinPython 3.13.0.0dotThe following packages are included in WinPython-64bit v3.13.0.0dot a2. Tools
Python packages
|
|
It seems that numpy is now smaller and linked to the mkl package in the latest C Gohlke release: https://github.com/cgohlke/numpy-mkl-wheels/releases/tag/v2024.1.3 Would this make it easier to have a WinPython+mkl release included going forward? |
In theory yes, if VirusTotal was happy with intel_openmp-2024.0.2-py2.py3-none-win_amd64.whl package. |
|
counting T-Tops: A100 / 4090 = 1200 Tops, 4070 = 450 Tops, 4060ti = 4070 laptop= 350 Tops, 4050 laptop = 50tops(?), M3 Max = 35 Tops/400Gbps, Zen 5 HawkPoint = 15 Tops / 135Gbps... bandwith / ram Size / Speed is a bottleneck future (mid 24): Zen 5 StrixPoint = 50(?) Tops, SnapDragon X elite = 45/75tops/135Gbps, noting xps14 comes with the copilot button and a rtx4050, xps16 up to rtx4070, amd RocM praised on Linux, Cuda on Windows an interesting article, even if biased, from Ampere CPO, summarised in the title: https://www.techradar.com/pro/for-many-ai-applications-gpus-are-compute-overkill-consuming-much-more-power-and-money-than-needed-how-ampere-computing-plans-to-ride-the-ai-wave So, the speach could be:
|
|
WinPython 2024-01 b2 (2024-01-27) Changes from WinPython 2024-01 b1:
Areas of particular interest for testers:
next build:
binaries at:
WinPython 3.11.7.0The following packages are included in WinPython-64bit v3.11.7.0 b2. Tools
Python packages
WinPython 3.12.1.0The following packages are included in WinPython-64bit v3.12.1.0 b2. Tools
Python packages
|

Target date: mid february 2024
Wanted:
Focus:
Exploratory:
Direction:
The text was updated successfully, but these errors were encountered: