bug: [v0.5.15] Running DeepSeek R1 error:"Failed to fetch"
Opened this issue · 0 comments
david-menloai commented
I have correctly installed the NVIDIA GPU driver version 551.23 and CUDA Toolkit 12.4. However, when I load the model, the software throws an error: "Failed to fetch." I have repeatedly lowered the values for Max Tokens and Number of GPU Layers, but the error persists.
My Jan version is 0.5.15, and the llama-cpp is "windows-amd64-avx2-cuda-12-0." The Windows version is Windows 10 Workstation 19045.5487. I would greatly appreciate any help in resolving this issue!
DeviceSpecs (1).txt
cortex (1).log
app (1).log
Note: here is the issue in the log
cortex.llamacpp\llama.cpp\ggml\src\ggml-cuda\template-instances../fattn-vec-f16.cuh:317: GGML_ASSERT(precision == GGML_PREC_DEFAULT) failed