Popular repositories Loading

1. bevformer-blackwell bevformer-blackwell Public

   Run BEVFormer (ECCV 2022) on PyTorch 2.8 + Blackwell B200; SimpleDDP fix for mmcv 1.x on PyTorch 2.x + Tensor Core analysis

   Shell 5 1

2. faster-whisper-rocm-strix-halo faster-whisper-rocm-strix-halo Public

   Drop-in recipe for running faster-whisper on AMD Strix Halo (Ryzen AI Max+ 395, gfx1151) with Ubuntu 26.04 + ROCm 7.2.2 — no source build required

   Python 3 1

3. dense-27b-31b-dgx-spark dense-27b-31b-dgx-spark Public

   Real-world benchmark of Qwen 3.5/3.6-27B and Gemma 4-31B Dense models on NVIDIA DGX Spark (GB10) with llama.cpp b8922.

   Shell 2

4. hip-vs-vulkan-evo-x2 hip-vs-vulkan-evo-x2 Public

   llama.cpp HIP vs Vulkan benchmark on AMD Strix Halo (gfx1151) with ROCm 7.2.2 — Ubuntu 26.04

   Python 2

5. gemma4-vs-qwen35-dgx-spark gemma4-vs-qwen35-dgx-spark Public

   Gemma 4 vs Qwen 3.5 MoE benchmark on DGX Spark with llama.cpp

   Python 1

6. qwen36-evo-x2 qwen36-evo-x2 Public

   Benchmarks comparing 3 Qwen 3.6-35B-A3B GGUF quantizations (bartowski Q4_K_M, Unsloth UD-Q4_K_M, MXFP4_MOE) on GMKtec EVO-X2 with llama.cpp Vulkan. Includes JCommonsenseQA (nothink + Thinking ON), ...

   Python 1