matt/ollama37
1
0
Fork 0
mirror of https://github.com/dogkeeper886/ollama37.git synced 2025-12-11 08:17:03 +00:00
Code Issues Packages Projects Releases Wiki Activity

int

Browse Source
This commit is contained in:
Michael Yang
2024-07-03 19:41:17 -07:00
parent e2c3f6b3e2
commit 0f1910129f
5 changed files with 42 additions and 77 deletions
Download Patch File Download Diff File Expand all files Collapse all files

23
server/sched.go
View File

@@ -5,9 +5,11 @@ import (
"errors"
"fmt"
"log/slog"
"os"
"reflect"
"runtime"
"sort"
"strconv"
"strings"
"sync"
"time"
@@ -59,11 +61,12 @@ var defaultParallel = 4
var ErrMaxQueue = fmt.Errorf("server busy, please try again. maximum pending requests exceeded")
func InitScheduler(ctx context.Context) *Scheduler {
maxQueue := envconfig.MaxQueue()
sched := &Scheduler{
pendingReqCh: make(chan *LlmRequest, envconfig.MaxQueuedRequests),
finishedReqCh: make(chan *LlmRequest, envconfig.MaxQueuedRequests),
expiredCh: make(chan *runnerRef, envconfig.MaxQueuedRequests),
unloadedCh: make(chan interface{}, envconfig.MaxQueuedRequests),
pendingReqCh: make(chan *LlmRequest, maxQueue),
finishedReqCh: make(chan *LlmRequest, maxQueue),
expiredCh: make(chan *runnerRef, maxQueue),
unloadedCh: make(chan interface{}, maxQueue),
loaded: make(map[string]*runnerRef),
newServerFn: llm.NewLlamaServer,
getGpuFn: gpu.GetGPUInfo,
@@ -126,7 +129,7 @@ func (s *Scheduler) processPending(ctx context.Context) {
slog.Debug("pending request cancelled or timed out, skipping scheduling")
continue
}
numParallel := envconfig.NumParallel
numParallel := envconfig.NumParallel()
// TODO (jmorganca): multimodal models don't support parallel yet
// see https://github.com/ollama/ollama/issues/4165
if len(pending.model.ProjectorPaths) > 0 && numParallel != 1 {
@@ -148,7 +151,7 @@ func (s *Scheduler) processPending(ctx context.Context) {
pending.useLoadedRunner(runner, s.finishedReqCh)
break
}
} else if envconfig.MaxRunners > 0 && loadedCount >= envconfig.MaxRunners {
} else if envconfig.MaxRunners() > 0 && loadedCount >= envconfig.MaxRunners() {
slog.Debug("max runners achieved, unloading one to make room", "runner_count", loadedCount)
runnerToExpire = s.findRunnerToUnload()
} else {
@@ -161,7 +164,7 @@ func (s *Scheduler) processPending(ctx context.Context) {
gpus = s.getGpuFn()
}
if envconfig.MaxRunners <= 0 {
if envconfig.MaxRunners() <= 0 {
// No user specified MaxRunners, so figure out what automatic setting to use
// If all GPUs have reliable free memory reporting, defaultModelsPerGPU * the number of GPUs
// if any GPU has unreliable free memory reporting, 1x the number of GPUs
@@ -173,11 +176,13 @@ func (s *Scheduler) processPending(ctx context.Context) {
}
}
if allReliable {
envconfig.MaxRunners = defaultModelsPerGPU * len(gpus)
// HACK
os.Setenv("OLLAMA_MAX_LOADED_MODELS", strconv.Itoa(defaultModelsPerGPU*len(gpus)))
slog.Debug("updating default concurrency", "OLLAMA_MAX_LOADED_MODELS", envconfig.MaxRunners, "gpu_count", len(gpus))
} else {
// HACK
os.Setenv("OLLAMA_MAX_LOADED_MODELS", strconv.Itoa(len(gpus)))
slog.Info("one or more GPUs detected that are unable to accurately report free memory - disabling default concurrency")
envconfig.MaxRunners = len(gpus)
}
}