Merge pull request #9824 from ollama/mxyng/sched
conditionally enable parallel pipelines
This commit is contained in:
commit
021dcf089d
@ -373,7 +373,7 @@ func New(r *os.File, params ml.BackendParams) (ml.Backend, error) {
|
|||||||
(*C.ggml_backend_buffer_type_t)(unsafe.Pointer(&schedBufts[0])),
|
(*C.ggml_backend_buffer_type_t)(unsafe.Pointer(&schedBufts[0])),
|
||||||
C.int(len(schedBackends)),
|
C.int(len(schedBackends)),
|
||||||
C.size_t(maxGraphNodes),
|
C.size_t(maxGraphNodes),
|
||||||
true,
|
C._Bool(len(gpus) > 1 && slices.Contains(gpus, output.d)),
|
||||||
),
|
),
|
||||||
input: deviceBufferTypes[input.d],
|
input: deviceBufferTypes[input.d],
|
||||||
output: deviceBufferTypes[output.d],
|
output: deviceBufferTypes[output.d],
|
||||||
|
Loading…
x
Reference in New Issue
Block a user