Tests
Suites
Latest Results
Search
Register
Login
Popular Tests
Timed Linux Kernel Compilation
SVT-AV1
7-Zip Compression
Stockfish
x265
FFmpeg
Newest Tests
LiteRT
WarpX
Epoch
Valkey
Whisperfile
XNNPACK
Recently Updated Tests
ASTC Encoder
SVT-AV1
Unvanquished
Primesieve
XNNPACK
oneDNN
New & Recently Updated Tests
Recently Updated Suites
Database Test Suite
Machine Learning
Steam
New & Recently Updated Suites
Component Benchmarks
CPUs / Processors
GPUs / Graphics
OpenGL
Disks / Storage
Motherboards
File-Systems
Operating Systems
OpenBenchmarking.org
Corporate / Organization Info
Bug Reports / Feature Requests
Llama.cpp 1.1.0
pts/llama-cpp-1.1.0
- 02 June 2024 -
Update against Llama.cpp upstream, switch to Llama 3 model.
downloads.xml
<?xml version="1.0"?> <!--Phoronix Test Suite v10.8.5--> <PhoronixTestSuite> <Downloads> <Package> <URL>https://github.com/ggerganov/llama.cpp/archive/refs/tags/b3067.tar.gz</URL> <MD5>951dd9ce10456799db6ad24d0d5b65bd</MD5> <SHA256>440319d7ace91c6ea009fb37f723150a6be7cadbb27110f86744896e278fbb86</SHA256> <FileName>llama.cpp-b3067.tar.gz</FileName> <FileSize>20214347</FileSize> </Package> <Package> <URL>https://huggingface.co/lmstudio-community/Meta-Llama-3-8B-Instruct-GGUF/resolve/0910a3e69201d274d4fd68e89448114cd78e4c82/Meta-Llama-3-8B-Instruct-Q8_0.gguf?download=true</URL> <MD5>75bb6264a938d98d9a0b3af9dced2985</MD5> <SHA256>4514087a6e21a05906121ebbe22e5f8610eab5e5db561d3a1be20c7fe43167c8</SHA256> <FileName>Meta-Llama-3-8B-Instruct-Q8_0.gguf</FileName> <FileSize>8540770496</FileSize> <Optional>TRUE</Optional> </Package> </Downloads> </PhoronixTestSuite>
install.sh
#!/bin/bash tar -xf llama.cpp-b3067.tar.gz cd llama.cpp-b3067 make -j LLAMA_OPENBLAS=1 echo $? > ~/install-exit-status echo "#!/bin/sh cd llama.cpp-b3067 ./main \$@ -p \"Building a website can be done in 10 simple steps:\" -n 512 -e -t \$NUM_CPU_PHYSICAL_CORES > \$LOG_FILE 2>&1 echo \$? > ~/test-exit-status" > ~/llama-cpp chmod +x ~/llama-cpp
results-definition.xml
<?xml version="1.0"?> <!--Phoronix Test Suite v10.8.5--> <PhoronixTestSuite> <ResultsParser> <OutputTemplate>llama_print_timings: eval time = 38874.00 ms / 511 runs ( 76.07 ms per token, #_RESULT_# tokens per second)</OutputTemplate> <LineHint>tokens per second</LineHint> <ResultBeforeString>tokens</ResultBeforeString> </ResultsParser> </PhoronixTestSuite>
test-definition.xml
<?xml version="1.0"?> <!--Phoronix Test Suite v10.8.5--> <PhoronixTestSuite> <TestInformation> <Title>Llama.cpp</Title> <AppVersion>b3067</AppVersion> <Description>Llama.cpp is a port of Facebook's LLaMA model in C/C++ developed by Georgi Gerganov. Llama.cpp allows the inference of LLaMA and other supported models in C/C++. For CPU inference Llama.cpp supports AVX2/AVX-512, ARM NEON, and other modern ISAs along with features like OpenBLAS usage.</Description> <ResultScale>Tokens Per Second</ResultScale> <Proportion>HIB</Proportion> <TimesToRun>3</TimesToRun> </TestInformation> <TestProfile> <Version>1.1.0</Version> <SupportedPlatforms>Linux</SupportedPlatforms> <SoftwareType>Utility</SoftwareType> <TestType>System</TestType> <License>Free</License> <ExternalDependencies>build-utilities, blas-development</ExternalDependencies> <InstallRequiresInternet>TRUE</InstallRequiresInternet> <EnvironmentSize>58700</EnvironmentSize> <ProjectURL>https://github.com/ggerganov/llama.cpp/</ProjectURL> <RepositoryURL>https://github.com/ggerganov/llama.cpp</RepositoryURL> <Maintainer>Michael Larabel</Maintainer> <SystemDependencies>pkgconf</SystemDependencies> </TestProfile> <TestSettings> <Option> <DisplayName>Model</DisplayName> <Identifier>model</Identifier> <ArgumentPrefix>-m ../</ArgumentPrefix> <Menu> <Entry> <Name>Meta-Llama-3-8B-Instruct-Q8_0.gguf</Name> <Value>Meta-Llama-3-8B-Instruct-Q8_0.gguf</Value> </Entry> </Menu> </Option> </TestSettings> </PhoronixTestSuite>