highperfocused/ollama
1
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2025-06-14 18:40:55 +02:00
Code Issues Packages Projects Releases Wiki Activity

config: allow setting context length through env var (#8938)

Browse Source 
* envconfig: allow setting context length through env var
This commit is contained in:
Parth Sareen 2025-02-24 13:26:35 -08:00 committed by GitHub
parent 4604b10306
commit 314573bfe8
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
4 changed files with 23 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
api/types.go
View File

@ -10,6 +10,8 @@ import (
"strconv" "strconv"
"strings" "strings"
"time" "time"
"github.com/ollama/ollama/envconfig"
) )
// StatusError is an error with an HTTP status code and message. // StatusError is an error with an HTTP status code and message.
@ -609,7 +611,7 @@ func DefaultOptions() Options {
Runner: Runner{ Runner: Runner{
// options set when the model is loaded // options set when the model is loaded
NumCtx: 2048, NumCtx: int(envconfig.ContextLength()),
NumBatch: 512, NumBatch: 512,
NumGPU: -1, // -1 here indicates that NumGPU should be set dynamically NumGPU: -1, // -1 here indicates that NumGPU should be set dynamically
NumThread: 0, // let the runtime decide NumThread: 0, // let the runtime decide

3
envconfig/config.go
View File

@ -167,6 +167,8 @@ var (
MultiUserCache = Bool("OLLAMA_MULTIUSER_CACHE") MultiUserCache = Bool("OLLAMA_MULTIUSER_CACHE")
// Enable the new Ollama engine // Enable the new Ollama engine
NewEngine = Bool("OLLAMA_NEW_ENGINE") NewEngine = Bool("OLLAMA_NEW_ENGINE")
// ContextLength sets the default context length
ContextLength = Uint("OLLAMA_CONTEXT_LENGTH", 2048)
) )
func String(s string) func() string { func String(s string) func() string {
@ -252,6 +254,7 @@ func AsMap() map[string]EnvVar {
"OLLAMA_ORIGINS": {"OLLAMA_ORIGINS", AllowedOrigins(), "A comma separated list of allowed origins"}, "OLLAMA_ORIGINS": {"OLLAMA_ORIGINS", AllowedOrigins(), "A comma separated list of allowed origins"},
"OLLAMA_SCHED_SPREAD": {"OLLAMA_SCHED_SPREAD", SchedSpread(), "Always schedule model across all GPUs"}, "OLLAMA_SCHED_SPREAD": {"OLLAMA_SCHED_SPREAD", SchedSpread(), "Always schedule model across all GPUs"},
"OLLAMA_MULTIUSER_CACHE": {"OLLAMA_MULTIUSER_CACHE", MultiUserCache(), "Optimize prompt caching for multi-user scenarios"}, "OLLAMA_MULTIUSER_CACHE": {"OLLAMA_MULTIUSER_CACHE", MultiUserCache(), "Optimize prompt caching for multi-user scenarios"},
"OLLAMA_CONTEXT_LENGTH": {"OLLAMA_CONTEXT_LENGTH", ContextLength(), "Context length to use unless otherwise specified (default: 2048)"},
"OLLAMA_NEW_ENGINE": {"OLLAMA_NEW_ENGINE", NewEngine(), "Enable the new Ollama engine"}, "OLLAMA_NEW_ENGINE": {"OLLAMA_NEW_ENGINE", NewEngine(), "Enable the new Ollama engine"},
// Informational // Informational

16
envconfig/config_test.go
View File

@ -272,3 +272,19 @@ func TestVar(t *testing.T) {
}) })
} }
} }
func TestContextLength(t *testing.T) {
cases := map[string]uint{
"": 2048,
"4096": 4096,
}
for k, v := range cases {
t.Run(k, func(t *testing.T) {
t.Setenv("OLLAMA_CONTEXT_LENGTH", k)
if i := ContextLength(); i != v {
t.Errorf("%s: expected %d, got %d", k, v, i)
}
})
}
}

1
llm/memory_test.go
View File

@ -17,6 +17,7 @@ import (
func TestEstimateGPULayers(t *testing.T) { func TestEstimateGPULayers(t *testing.T) {
t.Setenv("OLLAMA_DEBUG", "1") t.Setenv("OLLAMA_DEBUG", "1")
t.Setenv("OLLAMA_KV_CACHE_TYPE", "") // Ensure default f16 t.Setenv("OLLAMA_KV_CACHE_TYPE", "") // Ensure default f16
t.Setenv("OLLAMA_CONTEXT_LENGTH", "2048")
modelName := "dummy" modelName := "dummy"
f, err := os.CreateTemp(t.TempDir(), modelName) f, err := os.CreateTemp(t.TempDir(), modelName)