rtx5070

Dashboard for AI Studio, Open Source Continuous Inference | Deepseek-R1, Qwen2.5, Llama3.1 | 4xRTX-5090 inside PRU2500, 2xH100 inside PRU2500, 8xMI210 in SuperMicro

AI-generated RTX 5090 XBAR overclocking tool with response code for related branches/discussions. Tested only on the author's configuration. For other setups, follow the docs and use an AI agent to adapt it to your machine. Use at your own risk.

Prebuilt spconv v2.3.8 wheels for CUDA 12.8 / 13.0 with native Blackwell (RTX 50-series, sm_120) kernels — for default PyPI torch (cu130) or torch +cu128

Experimental Windows MSLK build for NVIDIA Blackwell (SM120) GPUs. Includes a validated Triton Split-K adjustment for head_dim=512 workloads used by xFormers.

Experimental xFormers + MSLK builds validated on NVIDIA Blackwell (SM120) GPUs, CUDA 12.8, PyTorch 2.11, and RTX 5070 Laptop hardware.

Evaluation, latency benchmarks, and voice cloning lab for OmniVoice zero-shot TTS on local desktop GPUs.