mirrors/ollama-for-amd
1
0
Fork 0
mirror of https://github.com/likelovewant/ollama-for-amd.git synced 2025-12-23 23:18:26 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'ollama:main' into main

Browse Source
This commit is contained in:
likelovewant
2024-05-29 19:33:39 +08:00
committed by GitHub
parent 2a80d6f743 646371f56d
commit cafde1f8ce
13 changed files with 966 additions and 130 deletions
Download Patch File Download Diff File Expand all files Collapse all files

30
llm/generate/gen_linux.sh
View File

@@ -215,6 +215,36 @@ if [ -z "${OLLAMA_SKIP_CUDA_GENERATE}" -a -d "${CUDA_LIB_DIR}" ]; then
fi
if [ -z "${ONEAPI_ROOT}" ]; then
# Try the default location in case it exists
ONEAPI_ROOT=/opt/intel/oneapi
fi
if [ -d "${ONEAPI_ROOT}" ]; then
echo "OneAPI libraries detected - building dynamic OneAPI library"
init_vars
source ${ONEAPI_ROOT}/setvars.sh --force # set up environment variables for oneAPI
CC=icx
CMAKE_DEFS="${COMMON_CMAKE_DEFS} ${CMAKE_DEFS} -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx -DLLAMA_SYCL=ON -DLLAMA_SYCL_F16=OFF"
BUILD_DIR="../build/linux/${ARCH}/oneapi"
EXTRA_LIBS="-fsycl -Wl,-rpath,${ONEAPI_ROOT}/compiler/latest/lib,-rpath,${ONEAPI_ROOT}/mkl/latest/lib,-rpath,${ONEAPI_ROOT}/tbb/latest/lib,-rpath,${ONEAPI_ROOT}/compiler/latest/opt/oclfpga/linux64/lib -lOpenCL -lmkl_core -lmkl_sycl_blas -lmkl_intel_ilp64 -lmkl_tbb_thread -ltbb"
DEBUG_FLAGS="" # icx compiles with -O0 if we pass -g, so we must remove it
build
# copy oneAPI dependencies
for dep in $(ldd "${BUILD_DIR}/bin/ollama_llama_server" | grep "=>" | cut -f2 -d= | cut -f2 -d' ' | grep -e sycl -e mkl -e tbb); do
cp "${dep}" "${BUILD_DIR}/bin/"
done
cp "${ONEAPI_ROOT}/compiler/latest/lib/libOpenCL.so" "${BUILD_DIR}/bin/"
cp "${ONEAPI_ROOT}/compiler/latest/lib/libimf.so" "${BUILD_DIR}/bin/"
cp "${ONEAPI_ROOT}/compiler/latest/lib/libintlc.so.5" "${BUILD_DIR}/bin/"
cp "${ONEAPI_ROOT}/compiler/latest/lib/libirng.so" "${BUILD_DIR}/bin/"
cp "${ONEAPI_ROOT}/compiler/latest/lib/libpi_level_zero.so" "${BUILD_DIR}/bin/"
cp "${ONEAPI_ROOT}/compiler/latest/lib/libsvml.so" "${BUILD_DIR}/bin/"
cp "${ONEAPI_ROOT}/compiler/latest/lib/libur_loader.so.0" "${BUILD_DIR}/bin/"
compress
fi
if [ -z "${ROCM_PATH}" ]; then
# Try the default location in case it exists
ROCM_PATH=/opt/rocm

44
llm/generate/gen_windows.ps1
View File

@@ -300,6 +300,49 @@ function build_cuda() {
}
}
function build_oneapi() {
if ((-not "${env:OLLAMA_SKIP_CUDA_GENERATE}") -and ("${env:ONEAPI_ROOT}")) {
# Get oneAPI version
$script:ONEAPI_VERSION = icpx --version
$script:ONEAPI_VERSION = [regex]::Match($script:ONEAPI_VERSION, '(?<=oneAPI DPC\+\+/C\+\+ Compiler )(?<version>\d+\.\d+\.\d+)').Value
if ($null -ne $script:ONEAPI_VERSION) {
$script:ONEAPI_VARIANT = "_v" + $script:ONEAPI_VERSION
}
init_vars
$script:buildDir = "../build/windows/${script:ARCH}/oneapi$script:ONEAPI_VARIANT"
$script:distDir ="$script:DIST_BASE\oneapi$script:ONEAPI_VARIANT"
$script:cmakeDefs += @(
"-G", "MinGW Makefiles",
"-DLLAMA_SYCL=ON",
"-DCMAKE_C_COMPILER=icx",
"-DCMAKE_CXX_COMPILER=icx",
"-DCMAKE_BUILD_TYPE=Release"
)
Write-Host "Building oneAPI"
build
# Ninja doesn't prefix with config name
if ($null -ne $script:DUMPBIN) {
& "$script:DUMPBIN" /dependents "${script:buildDir}/bin/ollama_llama_server.exe" | Select-String ".dll"
}
sign
install
cp "${env:ONEAPI_ROOT}\compiler\latest\bin\libirngmd.dll" "${script:distDir}"
cp "${env:ONEAPI_ROOT}\compiler\latest\bin\libmmd.dll" "${script:distDir}"
cp "${env:ONEAPI_ROOT}\compiler\latest\bin\pi_level_zero.dll" "${script:distDir}"
cp "${env:ONEAPI_ROOT}\compiler\latest\bin\pi_unified_runtime.dll" "${script:distDir}"
cp "${env:ONEAPI_ROOT}\compiler\latest\bin\pi_win_proxy_loader.dll" "${script:distDir}"
cp "${env:ONEAPI_ROOT}\compiler\latest\bin\svml_dispmd.dll" "${script:distDir}"
cp "${env:ONEAPI_ROOT}\compiler\latest\bin\sycl7.dll" "${script:distDir}"
cp "${env:ONEAPI_ROOT}\mkl\latest\bin\mkl_core.2.dll" "${script:distDir}"
cp "${env:ONEAPI_ROOT}\mkl\latest\bin\mkl_sycl_blas.4.dll" "${script:distDir}"
cp "${env:ONEAPI_ROOT}\mkl\latest\bin\mkl_tbb_thread.2.dll" "${script:distDir}"
} else {
Write-Host "Skipping oneAPI generation step"
}
}
function build_rocm() {
if ((-not "${env:OLLAMA_SKIP_ROCM_GENERATE}") -and ("${env:HIP_PATH}")) {
$script:ROCM_VERSION=(get-item $env:HIP_PATH).Basename
@@ -367,6 +410,7 @@ if ($($args.count) -eq 0) {
build_cpu_avx
build_cpu_avx2
build_cuda
build_oneapi
build_rocm
}

16
llm/server.go
View File

@@ -24,9 +24,9 @@ import (
"golang.org/x/sync/semaphore"
"github.com/ollama/ollama/api"
"github.com/ollama/ollama/envconfig"
"github.com/ollama/ollama/format"
"github.com/ollama/ollama/gpu"
"github.com/ollama/ollama/envconfig"
)
type LlamaServer interface {
@@ -243,7 +243,7 @@ func NewLlamaServer(gpus gpu.GpuInfoList, model string, ggml *GGML, adapters, pr
gpuCount = 0
}
// Find an availableServers port, retry on each iterration in case the failure was a port conflict race
// Find an availableServers port, retry on each iteration in case the failure was a port conflict race
port := 0
if a, err := net.ResolveTCPAddr("tcp", "localhost:0"); err == nil {
var l *net.TCPListener
@@ -519,11 +519,13 @@ func (s *llmServer) Ping(ctx context.Context) error {
func (s *llmServer) WaitUntilRunning(ctx context.Context) error {
start := time.Now()
stallDuration := 60 * time.Second
stallTimer := time.Now().Add(stallDuration) // give up if we stall for
stallDuration := 5 * time.Minute // If no progress happens
finalLoadDuration := 5 * time.Minute // After we hit 100%, give the runner more time to come online
stallTimer := time.Now().Add(stallDuration) // give up if we stall
slog.Info("waiting for llama runner to start responding")
var lastStatus ServerStatus = -1
fullyLoaded := false
for {
select {
@@ -572,6 +574,10 @@ func (s *llmServer) WaitUntilRunning(ctx context.Context) error {
if priorProgress != s.loadProgress {
slog.Debug(fmt.Sprintf("model load progress %0.2f", s.loadProgress))
stallTimer = time.Now().Add(stallDuration)
} else if !fullyLoaded && int(s.loadProgress*100.0) >= 100 {
slog.Debug("model load completed, waiting for server to become available", "status", status.ToString())
stallTimer = time.Now().Add(finalLoadDuration)
fullyLoaded = true
}
time.Sleep(time.Millisecond * 250)
continue
@@ -756,7 +762,7 @@ func (s *llmServer) Completion(ctx context.Context, req CompletionRequest, fn fu
var c completion
if err := json.Unmarshal(evt, &c); err != nil {
return fmt.Errorf("error unmarshaling llm prediction response: %v", err)
return fmt.Errorf("error unmarshalling llm prediction response: %v", err)
}
switch {