matt/ollama37
1
0
Fork 0
mirror of https://github.com/dogkeeper886/ollama37.git synced 2025-12-10 15:57:04 +00:00
Code Issues Packages Projects Releases Wiki Activity

Only set default keep_alive on initial model load

Browse Source 
This change fixes the handling of keep_alive so that if client
request omits the setting, we only set this on initial load.  Once
the model is loaded, if new requests leave this unset, we'll keep
whatever keep_alive was there.
This commit is contained in:
Daniel Hiltgen
2024-07-02 15:12:43 -07:00
parent ccd7785859
commit 955f2a4e03
5 changed files with 70 additions and 71 deletions
Download Patch File Download Diff File Expand all files Collapse all files

31
envconfig/config.go
View File

@@ -4,12 +4,14 @@ import (
"errors"
"fmt"
"log/slog"
"math"
"net"
"os"
"path/filepath"
"runtime"
"strconv"
"strings"
"time"
)
type OllamaHost struct {
@@ -34,7 +36,7 @@ var (
// Set via OLLAMA_HOST in the environment
Host *OllamaHost
// Set via OLLAMA_KEEP_ALIVE in the environment
KeepAlive string
KeepAlive time.Duration
// Set via OLLAMA_LLM_LIBRARY in the environment
LLMLibrary string
// Set via OLLAMA_MAX_LOADED_MODELS in the environment
@@ -132,6 +134,7 @@ func init() {
NumParallel = 0 // Autoselect
MaxRunners = 0 // Autoselect
MaxQueuedRequests = 512
KeepAlive = 5 * time.Minute
LoadConfig()
}
@@ -266,7 +269,10 @@ func LoadConfig() {
}
}
KeepAlive = clean("OLLAMA_KEEP_ALIVE")
ka := clean("OLLAMA_KEEP_ALIVE")
if ka != "" {
loadKeepAlive(ka)
}
var err error
ModelsDir, err = getModelsDir()
@@ -344,3 +350,24 @@ func getOllamaHost() (*OllamaHost, error) {
Port: port,
}, nil
}
func loadKeepAlive(ka string) {
v, err := strconv.Atoi(ka)
if err != nil {
d, err := time.ParseDuration(ka)
if err == nil {
if d < 0 {
KeepAlive = time.Duration(math.MaxInt64)
} else {
KeepAlive = d
}
}
} else {
d := time.Duration(v) * time.Second
if d < 0 {
KeepAlive = time.Duration(math.MaxInt64)
} else {
KeepAlive = d
}
}
}

17
envconfig/config_test.go
View File

@@ -2,8 +2,10 @@ package envconfig
import (
"fmt"
"math"
"net"
"testing"
"time"
"github.com/stretchr/testify/assert"
"github.com/stretchr/testify/require"
@@ -23,6 +25,21 @@ func TestConfig(t *testing.T) {
t.Setenv("OLLAMA_FLASH_ATTENTION", "1")
LoadConfig()
require.True(t, FlashAttention)
t.Setenv("OLLAMA_KEEP_ALIVE", "")
LoadConfig()
require.Equal(t, 5*time.Minute, KeepAlive)
t.Setenv("OLLAMA_KEEP_ALIVE", "3")
LoadConfig()
require.Equal(t, 3*time.Second, KeepAlive)
t.Setenv("OLLAMA_KEEP_ALIVE", "1h")
LoadConfig()
require.Equal(t, 1*time.Hour, KeepAlive)
t.Setenv("OLLAMA_KEEP_ALIVE", "-1s")
LoadConfig()
require.Equal(t, time.Duration(math.MaxInt64), KeepAlive)
t.Setenv("OLLAMA_KEEP_ALIVE", "-1")
LoadConfig()
require.Equal(t, time.Duration(math.MaxInt64), KeepAlive)
}
func TestClientFromEnvironment(t *testing.T) {