From 0c1041ad851d2ce3dc4d74d5fedb1af759956688 Mon Sep 17 00:00:00 2001 From: Bruce MacDonald Date: Thu, 27 Feb 2025 16:41:20 -0800 Subject: [PATCH] runner: default to greedy sampler for performance (#9407) As are adding support for weighted sampling we have seen some performance regressions, bypassing the sampler logic for now and defaulting to greedy until we can benchmark the new sampler logic. --- runner/ollamarunner/runner.go | 14 +------------- 1 file changed, 1 insertion(+), 13 deletions(-) diff --git a/runner/ollamarunner/runner.go b/runner/ollamarunner/runner.go index b39d747f8..6b4c7be07 100644 --- a/runner/ollamarunner/runner.go +++ b/runner/ollamarunner/runner.go @@ -575,23 +575,11 @@ func (s *Server) completion(w http.ResponseWriter, r *http.Request) { return } - sampler, err := sample.NewSampler( - req.Temperature, - req.TopK, - req.TopP, - req.MinP, - req.Seed, - ) - if err != nil { - http.Error(w, fmt.Sprintf("Failed to create sampler: %v", err), http.StatusInternalServerError) - return - } - seq, err := s.NewSequence(req.Prompt, req.Images, NewSequenceParams{ numPredict: req.NumPredict, stop: req.Stop, numKeep: int32(req.NumKeep), - sampler: sampler, + sampler: sample.Greedy(), // TODO: add support for different samplers when performance is optimized embedding: false, }) if err != nil {