whisper : fix max_tokens skipping remaining audio (#3798)

* whisper: fix max_tokens skipping remaining audio

* add PR reference comment as suggested

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

* fix(ci): enable artifact overwrite

bindings/go/pkg/whisper/context_test.go

@@ -2,6 +2,7 @@ package whisper_test
 
 import (
 	"os"
+	"strings"
 	"testing"
 
 	"github.com/ggerganov/whisper.cpp/bindings/go/pkg/whisper"
@@ -92,6 +93,53 @@ func TestProcess(t *testing.T) {
 	assert.NoError(err)
 }
 
+func TestProcessMaxTokensPerSegment(t *testing.T) {
+	assert := assert.New(t)
+
+	if _, err := os.Stat(ModelPath); os.IsNotExist(err) {
+		t.Skip("Skipping test, model not found:", ModelPath)
+	}
+
+	fh, err := os.Open(SamplePath)
+	assert.NoError(err)
+	defer fh.Close()
+
+	// Decode the WAV file - load the full buffer
+	dec := wav.NewDecoder(fh)
+	buf, err := dec.FullPCMBuffer()
+	assert.NoError(err)
+	assert.Equal(uint16(1), dec.NumChans)
+
+	data := buf.AsFloat32Buffer().Data
+
+	model, err := whisper.New(ModelPath)
+	assert.NoError(err)
+	assert.NotNil(model)
+	defer model.Close()
+
+	context, err := model.NewContext()
+	assert.NoError(err)
+
+	context.SetMaxTokensPerSegment(5)
+
+	err = context.Process(data, nil, nil, nil)
+	assert.NoError(err)
+
+	var text strings.Builder
+	nSegments := 0
+	for {
+		segment, err := context.NextSegment()
+		if err != nil {
+			break
+		}
+		nSegments++
+		text.WriteString(segment.Text)
+	}
+
+	assert.Greater(nSegments, 1)
+	assert.Contains(text.String(), "country")
+}
+
 func TestDetectedLanguage(t *testing.T) {
 	assert := assert.New(t)