Feature/modernize pytorch

[eoffermann]

Modernize PyTorch → 2.6 (+ CUDA 12.4) & Add CPU‑Only Support

This PR upgrades the entire PyTorch stack to the current 2.6 release, switches the CUDA wheels to 12.4, and introduces a standalone requirements‑cpu.txt for environments without a GPU. Along the way it fixes a few compatibility breaks that surfaced after the upgrade and removes an accidental binary blob.

---

✨ What’s new

Area	Change
Dependencies	* GPU path — requirements.txt now pins torch/torchvision/torchaudio 2.6.0+cu124 and bumps pytorch‑lightning to 2.5.1.* CPU path — new requirements‑cpu.txt with the matching CPU wheels.* Added in‑file section headers for clarity and trimmed unused version specifiers.
Code fixes	* gen_wav.py — passes the new weights_only=False flag required by PyTorch ≥2.5 when loading checkpoints.* ldm/models/diffusion/{ddpm,ddpm_audio}.py — import of rank_zero_only moved to the new namespace pytorch_lightning.utilities (old path was removed in PL 2.5).
Repo hygiene	* .gitignore now ignores stray .pth checkpoints.* Deleted the accidental useful_ckpts/.DS_Store binary.
Docs / comments	Updated inline comments so the “aligned with torch 2.6.0” lines reflect the real version instead of the old 1.13.

---

⚠️ Breaking / noteworthy impacts

• Python ≥3.9 is required by PyTorch 2.6 and PL 2.5.1.

• The rank_zero_only import path change means plugins or downstream forks must update their imports if they referenced the old location.

• Any local environment caches need a full pip install -r requirements.txt (or requirements‑cpu.txt) after the merge.

---

🔬 How I tested

• Fresh GPU and CPU virtual‑env installs on Linux and macOS (Apple Silicon in CPU mode) complete without conflicts.

• gen_wav.py end‑to‑end inference runs on both CUDA 12.4 and CPU, generating identical audio outputs relative to pre‑upgrade samples.

• Unit tests for the diffusion modules pass with PL 2.5.1 and the new import paths.

---

🚀 Upgrade guide

GPU build (CUDA 12.4)

python -m venv .venv && source .venv/bin/activate
pip install --upgrade pip
pip install -r requirements.txt

– or –

CPU build

python -m venv .venv && source .venv/bin/activate
pip install --upgrade pip
pip install -r requirements‑cpu.txt

---