Back to registry
CPU4GB VRAM

arnav080/llama-cpp-legacy-test

Stress test recipe simulating older llama.cpp configurations with flash attention, custom KV cache types, and NUMA settings

Configuration Specifications

Base Modelhuggingface:Qwen/Qwen2.5-0.5B-Instruct
Enginellama.cpp
QuantizationQ4_K_M
Platformcpu
VRAM Required4GB minimum

Telemetry Benchmarks

No runs recorded yet

Be the first to benchmark this recipe! Run the CLI command from your terminal:

bloc run arnav080/llama-cpp-legacy-test
llama-cpp-legacy-test.yaml
Loading workspace editor...