From eaa1d17756c7b41493532c5f4ea1ebfaf677d7aa Mon Sep 17 00:00:00 2001 From: Anish Raj Date: Sun, 31 May 2026 02:27:34 +0530 Subject: [PATCH 1/2] docs(phase-00/01): note Python 3.14 lacks PyTorch CUDA wheels (#192) --- phases/00-setup-and-tooling/01-dev-environment/docs/en.md | 2 ++ 1 file changed, 2 insertions(+) diff --git a/phases/00-setup-and-tooling/01-dev-environment/docs/en.md b/phases/00-setup-and-tooling/01-dev-environment/docs/en.md index 0c6a82a69..b73afde5d 100644 --- a/phases/00-setup-and-tooling/01-dev-environment/docs/en.md +++ b/phases/00-setup-and-tooling/01-dev-environment/docs/en.md @@ -123,6 +123,8 @@ nvidia-smi uv pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu124 ``` +> **On Python 3.14?** PyTorch doesn't ship CUDA wheels for the newest Python release right away, so the command above can't find a matching GPU build and you fall back to CPU-only torch. Use Python 3.11–3.13 for GPU support until the CUDA wheels catch up to 3.14. + ```python import torch print(f"CUDA available: {torch.cuda.is_available()}") From c91a8a07f44a8b72961b4dc2c445a68b68bb9705 Mon Sep 17 00:00:00 2001 From: Anish Raj Date: Sun, 31 May 2026 02:36:29 +0530 Subject: [PATCH 2/2] docs(phase-00/01): point Python 3.14 users to cu128/cu126 wheels --- phases/00-setup-and-tooling/01-dev-environment/docs/en.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/phases/00-setup-and-tooling/01-dev-environment/docs/en.md b/phases/00-setup-and-tooling/01-dev-environment/docs/en.md index b73afde5d..40c0b219c 100644 --- a/phases/00-setup-and-tooling/01-dev-environment/docs/en.md +++ b/phases/00-setup-and-tooling/01-dev-environment/docs/en.md @@ -123,7 +123,7 @@ nvidia-smi uv pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu124 ``` -> **On Python 3.14?** PyTorch doesn't ship CUDA wheels for the newest Python release right away, so the command above can't find a matching GPU build and you fall back to CPU-only torch. Use Python 3.11–3.13 for GPU support until the CUDA wheels catch up to 3.14. +> **On Python 3.14?** The `cu124` index above doesn't have 3.14 wheels, so the install falls back to CPU-only torch. The newer CUDA builds do ship 3.14 wheels — use `--index-url https://download.pytorch.org/whl/cu128` (or `cu126`) instead. If you'd rather stay on `cu124`, use Python 3.11–3.13. ```python import torch