v1.3.1 - Turing GPU Hotfix & Linux Support

This release brings a critical stability hotfix for pre-Ampere NVIDIA GPUs and introduces comprehensive Linux/Ubuntu installation instructions.

🛠️ Changes & Improvements:
Intellectual GPU Detection (Critical Hotfix): Added automatic CUDA compute capability verification. SageAttention is now safely auto-disabled for Turing (RTX 20xx / GTX 16xx) and older GPUs (compute capability < 8.0) when set to "auto".
Why: SageAttention does not officially support pre-Ampere GPUs. Running it on Turing with new PyTorch environments caused infinite Triton JIT compilation errors (MLIR type mismatch) and severe performance degradation (up to 2.0x speed drop). The node now automatically and smoothly falls back to stable native PyTorch SDPA.
🐧 Linux Setup Guide:** Added step-by-step setup guides for Linux/Ubuntu environments in both English and Russian READMEs.
🔍 Triton JIT Troubleshooting:** Documented solutions for common PyTorch Inductor compiler errors (such as PassManager::run failed under Ubuntu 24.04).

v1.3.0

This minor update introduces the TeaCache Version Selector, giving you full control over how timesteps are normalized and cached during generation. We have brought back the highly requested, ultra-fast SDE caching behavior as an optional legacy mode while preserving the mathematically precise mode for strict quality control.

🆕 What's New in v1.3.0:

⚡ TeaCache Version Selector (teacache_version)

You can now choose between two distinct caching behaviors directly in the AnimaTeaCache node:

• v1 (Legacy Fast) (Default):

  • How it works: Restores the legacy step normalizer (fixed at 1000.0) from version 1.2.0.
  • Best for: Instant 1.5×–2.0× speedup out-of-the-box on stochastic/SDE samplers (e.g., er_sde, sde gpu). It triggers highly aggressive caching from the very first steps, bypassing manual fine-tuning.
  • Note: May introduce minor structural changes or artifacts on Euler A.

• v2 (Standard Precise):

  • How it works: Employs mathematically precise, dynamic timestep normalization that adapts seamlessly to any sampler range (sigmas, 1000..0, or 1..0).
  • Best for: Samplers like euler_a and workflows requiring absolute geometric accuracy. It fully protects early structure-forming steps and relies on early_steps_factor and late_steps_factor for fine-tuning.

📦 Codebase Changes:

• Added teacache_version parameter to the Python node definitions and model patching core.
• Updated package versioning to v1.3.0 (__init__.py and UI configs).
• Updated comprehensive documentation in both English and Russian READMEs.

v1.2.1: BSS Premium UI & Runtime Widget Optimization

💎 What's New in v1.2.1

This update completely overhauls the user interface for the ANIMA_BOOSTER nodes, delivering a highly polished, premium, and distraction-free user experience in ComfyUI.

🌑 Integrated BSS Premium UI Theme

• High-Contrast Matte Design: Node bodies have been updated to a deep matte-black #0f0f0eff color scheme. This ensures exceptional contrast on any ComfyUI grid while seamlessly blending inputs and outputs into a monolithic block.
• Warm Gold Accents: All interactive elements—including active slider lines, toggles, input borders, and dropdown arrows—now feature an elegant golden #d4af37 / #e5c158 palette.
• Ergonomic Sliders: Slider interaction zones now span the full width of the widgets at their bottom edge, guaranteeing responsive click registration and smooth drag transitions.
• Branded Engraving: A clean and subtle golden BSS OPTIMIZED brand mark is now engraved in the bottom-right corner of each node.

🔄 Dynamic Runtime Widget Reactivity

• Reactive Layouts: Resolution presets in Anima Latent Image and modes in Anima TeaCache now toggle their associated input fields (e.g., width/height, early/late factors) dynamically in real-time.
• Robust Renderer Hooks: Rendering updates are bound to background drawing cycles, preventing common visual glitches where old widget styles would get stuck.
• Precise Formatting: Floating-point values for thresholds, factors, and percentages are now explicitly formatted to show exactly two decimal places (e.g., 0.15, 0.10).

🔇 Complete Tooltip Suppression

• All intrusive, overlapping help and default tooltips have been fully removed:
  • Removed all hardcoded tooltip descriptions from Python node definitions.
  • Implemented dynamic runtime suppression in the JS extension to completely disable the default LiteGraph tooltip property across all widgets, inputs, and outputs.

ANIMA_BOOSTER v1.2.0 — Stability & Performance Update

This major update focuses on robust stability, codebase refactoring, and fixing scaling bugs for stochastic/SDE samplers. We have eliminated unstable components and made the suite bulletproof for everyday generation.

🆕 What's New in v1.2.0:

1. 🐛 TeaCache Fixed for SDE/Stochastic Samplers (e.g., er_sde, sde gpu)

• The Issue: Stochastic samplers working on a sigma scale previously confused TeaCache's fixed threshold. This triggered aggressive caching on the very first step, resulting in fast generations but heavily distorted images covered in artifacts.
• The Solution: Implemented dynamic timestep scale auto-detection (st.max_t). TeaCache now mathematically adapts to any sampler and scheduler (sigmas, 1000..0, or 1..0). Early structural steps are fully protected, while late-stage detailing is safely cached. Enjoy perfect image quality with SDE samplers!

2. 💎 Safe One-Click JIT Compilation (torch.compile)

• Unstable AnimaTorchCompile node removed: The complex external compilation node was prone to PyTorch crashes (CUDA Graphs tensor overwrite errors).
• Integrated JIT Toggle: We integrated a safe, one-click torch_compile toggle directly into Anima Booster Loader and Checkpoint Loader. It runs on the stable inductor backend (default mode) without CUDA Graphs. Enjoy the same +20% to +40% speed boost with 100% stability!

3. 🗑️ Codebase Cleanup & Optimization

• Removed AnimaSparseAttention: Local sparse attention on blocks trained on Full Attention destroyed global image geometry and caused structural artifacts.
• Removed AnimaTorchCompile: Replaced by the native, JIT toggle in the model loaders.
• The package is now cleaner, lighter, and completely safe.

4. 📦 Graceful Degradation & Portable Windows Support

• All high-performance modules (like SageAttention) are now fully optional. If not installed, the loader will seamlessly fall back to PyTorch's native SDPA without throwing import errors.
• Windows/Portable Tip: Refer to the installation instructions in the README to download and install precompiled Triton and SageAttention binary wheels for Windows portable environments.

---

🎛️ Recommended Settings for Maximum Speed & Quality:

• Anima Booster Loader: Set sage_attention to auto and enable torch_compile. (Note: The first 2-3 generations will have a warm-up phase while PyTorch compiles the blocks).
• Anima TeaCache: Set threshold to 0.15 and keep adaptive ON.
• For SDE Samplers (like er_sde): Now fully compatible and artifact-free! If you want to push the speed further while maintaining great quality, try raising the TeaCache threshold to 0.22 - 0.25.