added testcase

* fix oneapi errors on windows 10

app/ollama_welcome.ps1

@@ -4,5 +4,5 @@ write-host "Welcome to Ollama!"
 write-host ""
 write-host "Run your first model:"
 write-host ""
-write-host "`tollama run llama2"
+write-host "`tollama run llama3"
 write-host ""

format/format.go

@@ -2,32 +2,41 @@ package format
 
 import (
 	"fmt"
-	"math"
 )
 
 const (
 	Thousand = 1000
 	Million  = Thousand * 1000
 	Billion  = Million * 1000
+	Trillion = Billion * 1000
 )
 
 func HumanNumber(b uint64) string {
 	switch {
+	case b >= Trillion:
+		number := float64(b) / Trillion
+		return fmt.Sprintf("%sT", DecimalPlace(number))
 	case b >= Billion:
 		number := float64(b) / Billion
-		if number == math.Floor(number) {
-			return fmt.Sprintf("%.0fB", number) // no decimals if whole number
-		}
-		return fmt.Sprintf("%.1fB", number) // one decimal if not a whole number
+		return fmt.Sprintf("%sB", DecimalPlace(number))
 	case b >= Million:
 		number := float64(b) / Million
-		if number == math.Floor(number) {
-			return fmt.Sprintf("%.0fM", number) // no decimals if whole number
-		}
-		return fmt.Sprintf("%.2fM", number) // two decimals if not a whole number
+		return fmt.Sprintf("%sM", DecimalPlace(number))
 	case b >= Thousand:
-		return fmt.Sprintf("%.0fK", float64(b)/Thousand)
+		number := float64(b) / Thousand
+		return fmt.Sprintf("%sK", DecimalPlace(number))
 	default:
 		return fmt.Sprintf("%d", b)
 	}
 }
+
+func DecimalPlace(number float64) string {
+	switch {
+	case number >= 100:
+		return fmt.Sprintf("%.0f", number)
+	case number >= 10:
+		return fmt.Sprintf("%.1f", number)
+	default:
+		return fmt.Sprintf("%.2f", number)
+	}
+}

format/format_test.go

@@ -13,14 +13,15 @@ func TestHumanNumber(t *testing.T) {
 
 	testCases := []testCase{
 		{0, "0"},
-		{1000000, "1M"},
+		{1000000, "1.00M"},
 		{125000000, "125M"},
-		{500500000, "500.50M"},
-		{500550000, "500.55M"},
-		{1000000000, "1B"},
-		{2800000000, "2.8B"},
-		{2850000000, "2.9B"},
-		{1000000000000, "1000B"},
+		{500500000, "500M"},
+		{500550000, "501M"},
+		{1000000000, "1.00B"},
+		{2800000000, "2.80B"},
+		{2850000000, "2.85B"},
+		{28550000000, "28.6B"},
+		{1000000000000, "1.00T"},
 	}
 
 	for _, tc := range testCases {

gpu/gpu.go

@@ -16,13 +16,12 @@ import (
 	"os"
 	"path/filepath"
 	"runtime"
-	"strconv"
 	"strings"
 	"sync"
 	"unsafe"
 
-	"github.com/ollama/ollama/format"
 	"github.com/ollama/ollama/envconfig"
+	"github.com/ollama/ollama/format"
 )
 
 type handles struct {
@@ -105,8 +104,6 @@ func initGPUHandles() *handles {
 	var cudartMgmtPatterns []string
 	var nvcudaMgmtName string
 	var nvcudaMgmtPatterns []string
-	var oneapiMgmtName string
-	var oneapiMgmtPatterns []string
 
 	tmpDir, _ := PayloadsDir()
 	switch runtime.GOOS {
@@ -118,8 +115,6 @@ func initGPUHandles() *handles {
 		// Aligned with driver, we can't carry as payloads
 		nvcudaMgmtName = "nvcuda.dll"
 		nvcudaMgmtPatterns = NvcudaWindowsGlobs
-		oneapiMgmtName = "ze_intel_gpu64.dll"
-		oneapiMgmtPatterns = OneapiWindowsGlobs
 	case "linux":
 		cudartMgmtName = "libcudart.so*"
 		if tmpDir != "" {
@@ -130,8 +125,6 @@ func initGPUHandles() *handles {
 		// Aligned with driver, we can't carry as payloads
 		nvcudaMgmtName = "libcuda.so*"
 		nvcudaMgmtPatterns = NvcudaLinuxGlobs
-		oneapiMgmtName = "libze_intel_gpu.so"
-		oneapiMgmtPatterns = OneapiLinuxGlobs
 	default:
 		return gpuHandles
 	}
@@ -159,17 +152,6 @@ func initGPUHandles() *handles {
 		}
 	}
 
-	oneapiLibPaths := FindGPULibs(oneapiMgmtName, oneapiMgmtPatterns)
-	if len(oneapiLibPaths) > 0 {
-		deviceCount, oneapi, libPath := LoadOneapiMgmt(oneapiLibPaths)
-		if oneapi != nil {
-			slog.Debug("detected Intel GPUs", "library", libPath, "count", deviceCount)
-			gpuHandles.oneapi = oneapi
-			gpuHandles.deviceCount = deviceCount
-			return gpuHandles
-		}
-	}
-
 	return gpuHandles
 }
 
@@ -245,18 +227,6 @@ func GetGPUInfo() GpuInfoList {
 			// TODO potentially sort on our own algorithm instead of what the underlying GPU library does...
 			resp = append(resp, gpuInfo)
 		}
-		if gpuHandles.oneapi != nil {
-			gpuInfo := GpuInfo{
-				Library: "oneapi",
-			}
-			C.oneapi_check_vram(*gpuHandles.oneapi, &memInfo)
-			var totalFreeMem float64 = float64(memInfo.free) * 0.95 // work-around: leave some reserve vram for mkl lib used in ggml-sycl backend.
-			memInfo.free = C.uint64_t(totalFreeMem)
-			gpuInfo.TotalMemory = uint64(memInfo.total)
-			gpuInfo.FreeMemory = uint64(memInfo.free)
-			gpuInfo.ID = strconv.Itoa(i)
-			resp = append(resp, gpuInfo)
-		}
 	}
 
 	// Then AMD