3 ms·
Anyone else getting gibberish when running unsloth/Qwen3.6-35B-A3B-GGUF:UD-IQ4_XS on CUDA (llama.cpp b8815)? UD-Q4_K_XL is fine, as is Vulkan in general.
by cpburns2009 6mo ago
Anyone else getting gibberish when running unsloth/Qwen3.6-35B-A3B-GGUF:UD-IQ4_XS on CUDA (llama.cpp b8815)? UD-Q4_K_XL is fine, as is Vulkan in general.
- cpburns2009 6mo agoApparently it's a known issue with CUDA 13.2 [1]. [1] https://huggingface.co/unsloth/Qwen3.6-35B-A3B-GGUF/discussions/3#69e14f5d66451286dfed7892 https://huggingface.co/unsloth/Qwen3.6-35B-A3B-GGUF/discussi...
- danielhanchen 6mo agoYes sadly CUDA 13.2 is broken - NVIDIA will push a fix in CUDA 13.3