pali112
/
ollama
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

prealloc

This commit is contained in:
Michael Yang 2025-11-16 18:49:48 -08:00
parent 974ae8ef84
commit 78a75a30d8
22 changed files with 86 additions and 92 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
.golangci.yaml
View File

@ -63,6 +63,11 @@ linters:
usestdlibvars:
http-method: false
http-status-code: false
exclusions:
rules:
- path: _test\.go
linters:
- prealloc
formatters:
enable:

6
api/types.go
View File

@ -309,9 +309,9 @@ func (tp ToolProperty) ToTypeScriptType() string {
return mapToTypeScriptType(tp.Type[0])
}
var types []string
for _, t := range tp.Type {
types = append(types, mapToTypeScriptType(t))
types := make([]string, len(tp.Type))
for i, t := range tp.Type {
types[i] = mapToTypeScriptType(t)
}
return strings.Join(types, " | ")
}

6
app/ui/ui.go
View File

@ -1730,8 +1730,8 @@ func supportsWebSearchTools(model string) bool {
// buildChatRequest converts store.Chat to api.ChatRequest
func (s *Server) buildChatRequest(chat *store.Chat, model string, think any, availableTools []map[string]any) (*api.ChatRequest, error) {
var msgs []api.Message
for _, m := range chat.Messages {
msgs := make([]api.Message, len(chat.Messages))
for i, m := range chat.Messages {
// Skip empty messages if present
if m.Content == "" && m.Thinking == "" && len(m.ToolCalls) == 0 && len(m.Attachments) == 0 {
continue
@ -1789,7 +1789,7 @@ func (s *Server) buildChatRequest(chat *store.Chat, model string, think any, ava
s.log().Debug("unknown message role", "role", m.Role)
}
msgs = append(msgs, apiMsg)
msgs[i] = apiMsg
}
var thinkValue *api.ThinkValue

7
cmd/interactive.go
View File

@ -576,9 +576,8 @@ func extractFileNames(input string) []string {
func extractFileData(input string) (string, []api.ImageData, error) {
filePaths := extractFileNames(input)
var imgs []api.ImageData
for _, fp := range filePaths {
imgs := make([]api.ImageData, len(filePaths))
for i, fp := range filePaths {
nfp := normalizeFilePath(fp)
data, err := getImageData(nfp)
if errors.Is(err, os.ErrNotExist) {
@ -591,7 +590,7 @@ func extractFileData(input string) (string, []api.ImageData, error) {
input = strings.ReplaceAll(input, "'"+nfp+"'", "")
input = strings.ReplaceAll(input, "'"+fp+"'", "")
input = strings.ReplaceAll(input, fp, "")
imgs = append(imgs, data)
imgs[i] = data
}
return strings.TrimSpace(input), imgs, nil
}

2
convert/convert_bert.go
View File

@ -137,7 +137,7 @@ func (p *bertModel) KV(t *Tokenizer) ggml.KV {
}
func (p *bertModel) Tensors(ts []Tensor) []*ggml.Tensor {
var out []*ggml.Tensor
out := make([]*ggml.Tensor, 0, len(ts))
for _, t := range ts {
if slices.Contains([]string{
"embeddings.position_ids",

8
convert/convert_commandr.go
View File

@ -44,14 +44,14 @@ func (p *commandrModel) KV(t *Tokenizer) ggml.KV {
}
func (p *commandrModel) Tensors(ts []Tensor) []*ggml.Tensor {
var out []*ggml.Tensor
for _, t := range ts {
out = append(out, &ggml.Tensor{
out := make([]*ggml.Tensor, len(ts))
for i, t := range ts {
out[i] = &ggml.Tensor{
Name: t.Name(),
Kind: t.Kind(),
Shape: t.Shape(),
WriterTo: t,
})
}
}
return out

8
convert/convert_gemma.go
View File

@ -43,18 +43,18 @@ func (p *gemmaModel) KV(t *Tokenizer) ggml.KV {
}
func (p *gemmaModel) Tensors(ts []Tensor) []*ggml.Tensor {
var out []*ggml.Tensor
for _, t := range ts {
out := make([]*ggml.Tensor, len(ts))
for i, t := range ts {
if !strings.HasPrefix(t.Name(), "v.") && strings.HasSuffix(t.Name(), "_norm.weight") {
t.SetRepacker(p.addOne)
}
out = append(out, &ggml.Tensor{
out[i] = &ggml.Tensor{
Name: t.Name(),
Kind: t.Kind(),
Shape: t.Shape(),
WriterTo: t,
})
}
}
return out

8
convert/convert_gemma2_adapter.go
View File

@ -22,8 +22,8 @@ func (p *gemma2Adapter) KV(baseKV ggml.KV) ggml.KV {
}
func (p *gemma2Adapter) Tensors(ts []Tensor) []*ggml.Tensor {
var out []*ggml.Tensor
for _, t := range ts {
out := make([]*ggml.Tensor, len(ts))
for i, t := range ts {
shape := t.Shape()
if (strings.HasSuffix(t.Name(), "weight.lora_a") && shape[0] > shape[1]) ||
(strings.HasSuffix(t.Name(), "weight.lora_b") && shape[0] < shape[1]) {
@ -31,12 +31,12 @@ func (p *gemma2Adapter) Tensors(ts []Tensor) []*ggml.Tensor {
t.SetRepacker(p.repack)
}
out = append(out, &ggml.Tensor{
out[i] = &ggml.Tensor{
Name: t.Name(),
Kind: t.Kind(),
Shape: t.Shape(),
WriterTo: t,
})
}
}
return out

8
convert/convert_llama.go
View File

@ -127,7 +127,7 @@ func (p *llamaModel) KV(t *Tokenizer) ggml.KV {
}
func (p *llamaModel) Tensors(ts []Tensor) []*ggml.Tensor {
var out []*ggml.Tensor
out := make([]*ggml.Tensor, 0, len(ts)+1)
if p.RopeScaling.factors != nil {
out = append(out, &ggml.Tensor{
@ -176,9 +176,9 @@ func (p *llamaModel) Replacements() []string {
}
func (p *llamaModel) repack(name string, data []float32, shape []uint64) ([]float32, error) {
var dims []int
for _, dim := range shape {
dims = append(dims, int(dim))
dims := make([]int, len(shape))
for i, dim := range shape {
dims[i] = int(dim)
}
var heads uint32

8
convert/convert_llama_adapter.go
View File

@ -30,8 +30,8 @@ func (p *llamaAdapter) KV(baseKV ggml.KV) ggml.KV {
}
func (p *llamaAdapter) Tensors(ts []Tensor) []*ggml.Tensor {
var out []*ggml.Tensor
for _, t := range ts {
out := make([]*ggml.Tensor, len(ts))
for i, t := range ts {
shape := t.Shape()
if (strings.HasSuffix(t.Name(), "weight.lora_a") && shape[0] > shape[1]) ||
(strings.HasSuffix(t.Name(), "weight.lora_b") && shape[0] < shape[1]) {
@ -41,12 +41,12 @@ func (p *llamaAdapter) Tensors(ts []Tensor) []*ggml.Tensor {
t.SetRepacker(p.repack)
}
out = append(out, &ggml.Tensor{
out[i] = &ggml.Tensor{
Name: t.Name(),
Kind: t.Kind(),
Shape: shape,
WriterTo: t,
})
}
}
return out

15
convert/convert_mistral.go
View File

@ -90,9 +90,8 @@ func (p *mistral3Model) KV(t *Tokenizer) ggml.KV {
}
func (p *mistral3Model) Tensors(ts []Tensor) []*ggml.Tensor {
var out []*ggml.Tensor
for _, t := range ts {
out := make([]*ggml.Tensor, len(ts))
for i, t := range ts {
if !strings.HasPrefix(t.Name(), "v.") {
if strings.HasSuffix(t.Name(), ".attn_q.weight") ||
strings.HasSuffix(t.Name(), ".attn_k.weight") {
@ -100,12 +99,12 @@ func (p *mistral3Model) Tensors(ts []Tensor) []*ggml.Tensor {
}
}
out = append(out, &ggml.Tensor{
out[i] = &ggml.Tensor{
Name: t.Name(),
Kind: t.Kind(),
Shape: t.Shape(),
WriterTo: t,
})
}
}
return out
@ -145,9 +144,9 @@ func (p *mistral3Model) Replacements() []string {
}
func (p *mistral3Model) repack(name string, data []float32, shape []uint64) ([]float32, error) {
var dims []int
for _, dim := range shape {
dims = append(dims, int(dim))
dims := make([]int, len(shape))
for i, dim := range shape {
dims[i] = int(dim)
}
var heads uint32

8
convert/convert_qwen2.go
View File

@ -49,14 +49,14 @@ func (q *qwen2Model) KV(t *Tokenizer) ggml.KV {
}
func (q *qwen2Model) Tensors(ts []Tensor) []*ggml.Tensor {
var out []*ggml.Tensor
for _, t := range ts {
out = append(out, &ggml.Tensor{
out := make([]*ggml.Tensor, len(ts))
for i, t := range ts {
out[i] = &ggml.Tensor{
Name: t.Name(),
Kind: t.Kind(),
Shape: t.Shape(),
WriterTo: t,
})
}
}
return out

2
convert/tokenizer_spm.go
View File

@ -82,7 +82,7 @@ func parseSentencePiece(fsys fs.FS) (*Vocabulary, error) {
content string
}
var ts []t
ts := make([]t, 0, len(atm))
for content, id := range atm {
ts = append(ts, t{id, content})
}

6
llama/llama.go
View File

@ -263,9 +263,9 @@ func LoadModelFromFile(modelPath string, params ModelParams) (*Model, error) {
cparams.use_mmap = C.bool(params.UseMmap)
cparams.vocab_only = C.bool(params.VocabOnly)
var devices []C.ggml_backend_dev_t
for _, llamaID := range params.Devices {
devices = append(devices, C.ggml_backend_dev_get(C.size_t(llamaID)))
devices := make([]C.ggml_backend_dev_t, len(params.Devices))
for i, llamaID := range params.Devices {
devices[i] = C.ggml_backend_dev_get(C.size_t(llamaID))
}
if len(devices) > 0 {
devices = append(devices, C.ggml_backend_dev_t(C.NULL))

12
middleware/openai_test.go
View File

@ -845,19 +845,17 @@ func TestListMiddleware(t *testing.T) {
resp := httptest.NewRecorder()
router.ServeHTTP(resp, req)
var expected, actual map[string]any
err := json.Unmarshal([]byte(tc.resp), &expected)
if err != nil {
var want, got map[string]any
if err := json.Unmarshal([]byte(tc.resp), &want); err != nil {
t.Fatalf("failed to unmarshal expected response: %v", err)
}
err = json.Unmarshal(resp.Body.Bytes(), &actual)
if err != nil {
if err := json.Unmarshal(resp.Body.Bytes(), &got); err != nil {
t.Fatalf("failed to unmarshal actual response: %v", err)
}
if !reflect.DeepEqual(expected, actual) {
t.Errorf("responses did not match\nExpected: %+v\nActual: %+v", expected, actual)
if diff := cmp.Diff(want, got); diff != "" {
t.Errorf("response does not match (-want +got):\n%s", diff)
}
}
}

10
ml/backend/ggml/ggml.go
View File

@ -178,14 +178,14 @@ func New(modelPath string, params ml.BackendParams) (ml.Backend, error) {
requiredMemory.CPU.Cache = make([]uint64, blocks+1)
// create list of buffer types for each gpu
var gpuDeviceBufferTypes []deviceBufferType
gpuDeviceBufferTypes := make([]deviceBufferType, len(gpus))
requiredMemory.GPUs = make([]ml.DeviceMemory, len(gpus))
for i, d := range gpus {
bt := C.ggml_backend_dev_buffer_type(d)
gpuDeviceBufferTypes = append(gpuDeviceBufferTypes, deviceBufferType{
gpuDeviceBufferTypes[i] = deviceBufferType{
d: d,
bts: append([]C.ggml_backend_buffer_type_t{bt}, cpuDeviceBufferType.bts...),
})
}
btDeviceMemory[bt] = &requiredMemory.GPUs[i]
requiredMemory.GPUs[i].Name = C.GoString(C.ggml_backend_dev_name(d))
@ -354,8 +354,8 @@ func New(modelPath string, params ml.BackendParams) (ml.Backend, error) {
deviceBufferTypes := make(map[C.ggml_backend_dev_t]C.ggml_backend_buffer_type_t)
// create backends and buffer types used for the compute graph scheduler
var schedBackends []C.ggml_backend_t
var schedBufts []C.ggml_backend_buffer_type_t
schedBackends := make([]C.ggml_backend_t, 0, len(cpus)+len(accels)+len(gpus))
schedBufts := make([]C.ggml_backend_buffer_type_t, 0, len(cpus)+len(accels)+len(gpus))
for _, d := range append(gpus, append(accels, cpus...)...) {
b := backends[d]
bt := C.ggml_backend_get_default_buffer_type(b)

2
ml/device.go
View File

@ -219,7 +219,7 @@ type BackendMemory struct {
}
func (m BackendMemory) LogValue() slog.Value {
var attrs []slog.Attr
attrs := make([]slog.Attr, 0, 2+len(m.GPUs))
if m.InputWeights != 0 {
attrs = append(attrs, slog.Any("InputWeights", m.InputWeights))
}

6
model/models/mllama/model_text.go
View File

@ -223,8 +223,8 @@ func (m *TextModel) Forward(ctx ml.Context, inputIDs, positionIDs, outputs, cros
}
func newTextModel(c fs.Config) *TextModel {
var decoderLayers []TextDecoderLayer
for i := range c.Uint("block_count") {
decoderLayers := make([]TextDecoderLayer, c.Uint("block_count"))
for i := range decoderLayers {
var textDecoderLayer TextDecoderLayer
if slices.Contains(c.Ints("attention.cross_attention_layers"), int32(i)) {
textDecoderLayer = &TextCrossAttentionDecoderLayer{}
@ -232,7 +232,7 @@ func newTextModel(c fs.Config) *TextModel {
textDecoderLayer = &TextSelfAttentionDecoderLayer{}
}
decoderLayers = append(decoderLayers, textDecoderLayer)
decoderLayers[i] = textDecoderLayer
}
return &TextModel{

25
openai/openai.go
View File

@ -377,20 +377,19 @@ func ToCompleteChunk(id string, r api.GenerateResponse) CompletionChunk {
// ToListCompletion converts an api.ListResponse to ListCompletion
func ToListCompletion(r api.ListResponse) ListCompletion {
var data []Model
for _, m := range r.Models {
data = append(data, Model{
Id: m.Name,
Object: "model",
Created: m.ModifiedAt.Unix(),
OwnedBy: model.ParseName(m.Name).Namespace,
})
}
return ListCompletion{
Object: "list",
Data: data,
c := ListCompletion{Object: "list"}
if len(r.Models) > 0 {
c.Data = make([]Model, len(r.Models))
for i, m := range r.Models {
c.Data[i] = Model{
Id: m.Name,
Object: "model",
Created: m.ModifiedAt.Unix(),
OwnedBy: model.ParseName(m.Name).Namespace,
}
}
}
return c
}
// ToEmbeddingList converts an api.EmbedResponse to EmbeddingList

12
runner/common/stop.go
View File

@ -39,21 +39,15 @@ func TruncateStop(pieces []string, stop string) ([]string, bool) {
joined = joined[:index]
// Split truncated string back into pieces of original lengths
lengths := make([]int, len(pieces))
for i, piece := range pieces {
lengths[i] = len(piece)
}
var result []string
result := make([]string, 0, len(pieces))
tokenTruncated := false
start := 0
for _, length := range lengths {
for _, piece := range pieces {
if start >= len(joined) {
break
}
end := start + length
end := start + len(piece)
if end > len(joined) {
end = len(joined)
tokenTruncated = true

6
server/create.go
View File

@ -460,8 +460,8 @@ func kvFromLayers(baseLayers []*layerGGML) (ggml.KV, error) {
}
func createModel(r api.CreateRequest, name model.Name, baseLayers []*layerGGML, config *ConfigV2, fn func(resp api.ProgressResponse)) (err error) {
var layers []Layer
for _, layer := range baseLayers {
layers := make([]Layer, len(baseLayers))
for i, layer := range baseLayers {
if layer.GGML != nil {
quantType := strings.ToUpper(cmp.Or(r.Quantize, r.Quantization))
if quantType != "" && layer.GGML.Name() == "gguf" && layer.MediaType == "application/vnd.ollama.image.model" {
@ -486,7 +486,7 @@ func createModel(r api.CreateRequest, name model.Name, baseLayers []*layerGGML,
config.FileType = cmp.Or(config.FileType, layer.GGML.KV().FileType().String())
config.ModelFamilies = append(config.ModelFamilies, layer.GGML.KV().Architecture())
}
layers = append(layers, layer.Layer)
layers[i] = layer.Layer
}
if r.Template != "" {

8
server/routes.go
View File

@ -826,9 +826,9 @@ func (s *Server) EmbeddingsHandler(c *gin.Context) {
return
}
var e []float64
for _, v := range embedding {
e = append(e, float64(v))
e := make([]float64, len(embedding))
for i, v := range embedding {
e[i] = float64(v)
}
resp := api.EmbeddingResponse{
@ -1211,7 +1211,7 @@ func (s *Server) ListHandler(c *gin.Context) {
return
}
models := []api.ListModelResponse{}
models := make([]api.ListModelResponse, 0, len(ms))
for n, m := range ms {
var cf ConfigV2