Get up and running with Kimi, GLM, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and other models.
some gocritic
Michael Yangc895a7d13f74
diff --git a/.golangci.yaml b/.golangci.yamlindex 7dec49ded..df966a162 100644--- a/.golangci.yaml+++ b/.golangci.yaml@@ -14,4 +14,6 @@ linters:# - goimports- misspell- nilerr+ - nolintlint+ - nosprintfhostport- unuseddiff --git a/api/types.go b/api/types.goindex 4195a7c58..230f58e8d 100644--- a/api/types.go+++ b/api/types.go@@ -306,7 +306,7 @@ type GenerateResponse struct {// Model is the model name that generated the response.Model string `json:"model"`- //CreatedAt is the timestamp of the response.+ // CreatedAt is the timestamp of the response.CreatedAt time.Time `json:"created_at"`// Response is the textual response itself.diff --git a/convert/llama.go b/convert/llama.goindex 7853c4cfb..b4211b02d 100644--- a/convert/llama.go+++ b/convert/llama.go@@ -119,11 +119,12 @@ func llamaRepack(name string, params *Params, data []float32, shape []uint64) ([}var heads int- if strings.HasSuffix(name, "attn_q.weight") {+ switch {+ case strings.HasSuffix(name, "attn_q.weight"):heads = params.AttentionHeads- } else if strings.HasSuffix(name, "attn_k.weight") {+ case strings.HasSuffix(name, "attn_k.weight"):heads = cmp.Or(params.KeyValHeads, params.AttentionHeads)- } else {+ default:return nil, fmt.Errorf("unknown tensor name: %s", name)}diff --git a/convert/safetensors.go b/convert/safetensors.goindex 69270b871..f45687f10 100644--- a/convert/safetensors.go+++ b/convert/safetensors.go@@ -120,7 +120,7 @@ func (m *SafetensorFormat) readTensors(fn string, offset uint64, params *Params)Name: name,Kind: kind,Offset: offset,- Shape: shape[:],+ Shape: shape,}t.WriterTo = safetensorWriterTo{diff --git a/convert/tokenizer.go b/convert/tokenizer.goindex efeb5491f..fd6df5f53 100644--- a/convert/tokenizer.go+++ b/convert/tokenizer.go@@ -85,11 +85,8 @@ func parseTokens(dirpath string) (pre string, tokens []Token, merges []string, esha256sum := sha256.New()for _, pt := range t.PreTokenizer.PreTokenizers {- switch pt.Type {- case "Split":- if pt.Pattern.Regex != "" {- sha256sum.Write([]byte(pt.Pattern.Regex))- }+ if pt.Type == "Split" && pt.Pattern.Regex != "" {+ sha256sum.Write([]byte(pt.Pattern.Regex))}}diff --git a/convert/torch.go b/convert/torch.goindex b7ae0f762..eef41a48f 100644--- a/convert/torch.go+++ b/convert/torch.go@@ -88,7 +88,7 @@ func (tf *TorchFormat) GetTensors(dirpath string, params *Params) ([]llm.Tensor,Name: ggufName,Kind: kind,Offset: offset, // calculate the offset- Shape: shape[:],+ Shape: shape,}tensor.WriterTo = torchWriterTo{diff --git a/envconfig/config.go b/envconfig/config.goindex 875c9039e..77e3e789d 100644--- a/envconfig/config.go+++ b/envconfig/config.go@@ -127,7 +127,7 @@ func LoadConfig() {var paths []stringfor _, root := range []string{filepath.Dir(appExe), cwd} {paths = append(paths,- filepath.Join(root),+ root,filepath.Join(root, "windows-"+runtime.GOARCH),filepath.Join(root, "dist", "windows-"+runtime.GOARCH),)diff --git a/llm/server.go b/llm/server.goindex 3af8a3294..f40278655 100644--- a/llm/server.go+++ b/llm/server.go@@ -104,21 +104,22 @@ func NewLlamaServer(gpus gpu.GpuInfoList, model string, ggml *GGML, adapters, prvar layers intlayers, estimatedVRAM, estimatedTotal = EstimateGPULayers(gpus, ggml, projectors, opts)- if gpus[0].Library == "metal" && estimatedVRAM > systemMemory {+ switch {+ case gpus[0].Library == "metal" && estimatedVRAM > systemMemory:// disable partial offloading when model is greater than total system memory as this// can lead to locking up the systemopts.NumGPU = 0- } else if gpus[0].Library != "metal" && layers == 0 {+ case gpus[0].Library != "metal" && layers == 0:// Don't bother loading into the GPU if no layers can fitcpuRunner = serverForCpu()gpuCount = 0- } else if opts.NumGPU < 0 && layers > 0 && gpus[0].Library != "cpu" {+ case opts.NumGPU < 0 && layers > 0 && gpus[0].Library != "cpu":opts.NumGPU = layers}}// Loop through potential servers- finalErr := fmt.Errorf("no suitable llama servers found")+ finalErr := errors.New("no suitable llama servers found")if len(adapters) > 1 {return nil, errors.New("ollama supports only one lora adapter, but multiple were provided")@@ -284,7 +285,7 @@ func NewLlamaServer(gpus gpu.GpuInfoList, model string, ggml *GGML, adapters, prserver := filepath.Join(dir, "ollama_llama_server")if runtime.GOOS == "windows" {- server = server + ".exe"+ server += ".exe"}// Detect tmp cleaners wiping out the file@@ -459,7 +460,7 @@ func (s *llmServer) getServerStatus(ctx context.Context) (ServerStatus, error) {resp, err := http.DefaultClient.Do(req)if err != nil {if errors.Is(err, context.DeadlineExceeded) {- return ServerStatusNotResponding, fmt.Errorf("server not responding")+ return ServerStatusNotResponding, errors.New("server not responding")}return ServerStatusError, fmt.Errorf("health resp: %w", err)}diff --git a/server/sched.go b/server/sched.goindex 46fe2f602..3694b4d02 100644--- a/server/sched.go+++ b/server/sched.go@@ -66,7 +66,7 @@ func (s *Scheduler) GetRunner(c context.Context, model *Model, opts api.Options,opts.NumCtx = 4}- opts.NumCtx = opts.NumCtx * envconfig.NumParallel+ opts.NumCtx *= envconfig.NumParallelreq := &LlmRequest{ctx: c,diff --git a/types/model/name_test.go b/types/model/name_test.goindex 26d70ef33..c88fffdb5 100644--- a/types/model/name_test.go+++ b/types/model/name_test.go@@ -325,7 +325,7 @@ func TestParseNameFromFilepath(t *testing.T) {filepath.Join("host:port", "namespace", "model", "tag"): {Host: "host:port", Namespace: "namespace", Model: "model", Tag: "tag"},filepath.Join("namespace", "model", "tag"): {},filepath.Join("model", "tag"): {},- filepath.Join("model"): {},+ "model": {},filepath.Join("..", "..", "model", "tag"): {},filepath.Join("", "namespace", ".", "tag"): {},filepath.Join(".", ".", ".", "."): {},