feat: extend model configuration for llama.cpp

Allow to set `mmap`, `mlock`, `frequency_penalty`, `tfz`,
`tensor_split`, `main_gpu`.

Makefile

@@ -3,7 +3,7 @@ GOTEST=$(GOCMD) test
 GOVET=$(GOCMD) vet
 BINARY_NAME=local-ai
 
-GOLLAMA_VERSION?=37ef81d01ae0848575e416e48b41d112ef0d520e
+GOLLAMA_VERSION?=9a7c56ad810067660cb06b26ccf8a3ac9dd2b21d
 GPT4ALL_REPO?=https://github.com/go-skynet/gpt4all
 GPT4ALL_VERSION?=f7498c9
 GOGGMLTRANSFORMERS_VERSION?=bd765bb6f3b38a63f915f3725e488aad492eedd4

api/config.go

@@ -33,6 +33,11 @@ type Config struct {
 	MirostatTAU    float64           `yaml:"mirostat_tau"`
 	Mirostat       int               `yaml:"mirostat"`
 	NGPULayers     int               `yaml:"gpu_layers"`
+	MMap           bool              `yaml:"mmap"`
+	MMlock         bool              `yaml:"mmlock"`
+
+	TensorSplit           string `yaml:"tensor_split"`
+	MainGPU               string `yaml:"main_gpu"`
 	ImageGenerationAssets string `yaml:"asset_dir"`
 
 	PromptCachePath string `yaml:"prompt_cache_path"`
@@ -53,6 +58,12 @@ type ConfigMerger struct {
 	sync.Mutex
 }
 
+func defaultConfig(modelFile string) *Config {
+	return &Config{
+		OpenAIRequest: defaultRequest(modelFile),
+	}
+}
+
 func NewConfigMerger() *ConfigMerger {
 	return &ConfigMerger{
 		configs: make(map[string]Config),
@@ -308,13 +319,11 @@ func readConfig(modelFile string, input *OpenAIRequest, cm *ConfigMerger, loader
 	var config *Config
 	cfg, exists := cm.GetConfig(modelFile)
 	if !exists {
-		config = &Config{
-			OpenAIRequest: defaultRequest(modelFile),
-			ContextSize:   ctx,
-			Threads:       threads,
-			F16:           f16,
-			Debug:         debug,
-		}
+		config = defaultConfig(modelFile)
+		config.ContextSize = ctx
+		config.Threads = threads
+		config.F16 = f16
+		config.Debug = debug
 	} else {
 		config = &cfg
 	}

api/openai.go

@@ -4,8 +4,8 @@ import (
 	"bufio"
 	"bytes"
 	"encoding/base64"
-	"errors"
 	"encoding/json"
+	"errors"
 	"fmt"
 	"io"
 	"io/ioutil"
@@ -125,6 +125,9 @@ type OpenAIRequest struct {
 	MirostatTAU float64 `json:"mirostat_tau" yaml:"mirostat_tau"`
 	Mirostat    int     `json:"mirostat" yaml:"mirostat"`
 
+	FrequencyPenalty float64 `json:"frequency_penalty" yaml:"frequency_penalty"`
+	TFZ              float64 `json:"tfz" yaml:"tfz"`
+
 	Seed int `json:"seed" yaml:"seed"`
 
 	// Image (not supported by OpenAI)
@@ -191,7 +194,7 @@ func completionEndpoint(cm *ConfigMerger, o *Option) func(c *fiber.Ctx) error {
 		}
 
 		if input.Stream {
-			if (len(config.PromptStrings) > 1) {
+			if len(config.PromptStrings) > 1 {
 				return errors.New("cannot handle more than 1 `PromptStrings` when `Stream`ing")
 			}
 

api/prediction.go

@@ -39,6 +39,12 @@ func defaultLLamaOpts(c Config) []llama.ModelOption {
 		llamaOpts = append(llamaOpts, llama.SetGPULayers(c.NGPULayers))
 	}
 
+	llamaOpts = append(llamaOpts, llama.SetMMap(c.MMap))
+	llamaOpts = append(llamaOpts, llama.SetMainGPU(c.MainGPU))
+	llamaOpts = append(llamaOpts, llama.SetTensorSplit(c.TensorSplit))
+	if c.Batch != 0 {
+		llamaOpts = append(llamaOpts, llama.SetNBatch(c.Batch))
+	}
 	return llamaOpts
 }
 
@@ -217,6 +223,15 @@ func buildLLamaPredictOptions(c Config, modelPath string) []llama.PredictOption
 		predictOptions = append(predictOptions, llama.SetSeed(c.Seed))
 	}
 
+	//predictOptions = append(predictOptions, llama.SetLogitBias(c.Seed))
+
+	predictOptions = append(predictOptions, llama.SetFrequencyPenalty(c.FrequencyPenalty))
+	predictOptions = append(predictOptions, llama.SetMlock(c.MMlock))
+	predictOptions = append(predictOptions, llama.SetMemoryMap(c.MMap))
+	predictOptions = append(predictOptions, llama.SetPredictionMainGPU(c.MainGPU))
+	predictOptions = append(predictOptions, llama.SetPredictionTensorSplit(c.TensorSplit))
+	predictOptions = append(predictOptions, llama.SetTailFreeSamplingZ(c.TFZ))
+
 	return predictOptions
 }
 

go.mod

@@ -9,7 +9,7 @@ require (
 	github.com/go-skynet/bloomz.cpp v0.0.0-20230529155654-1834e77b83fa
 	github.com/go-skynet/go-bert.cpp v0.0.0-20230531070950-0548994371f7
 	github.com/go-skynet/go-ggml-transformers.cpp v0.0.0-20230606131358-bd765bb6f3b3
-	github.com/go-skynet/go-llama.cpp v0.0.0-20230606152241-37ef81d01ae0
+	github.com/go-skynet/go-llama.cpp v0.0.0-20230607000523-9a7c56ad8100
 	github.com/gofiber/fiber/v2 v2.46.0
 	github.com/google/uuid v1.3.0
 	github.com/hashicorp/go-multierror v1.1.1
@@ -17,7 +17,7 @@ require (
 	github.com/mudler/go-stable-diffusion v0.0.0-20230605122230-d89260f598af
 	github.com/nomic-ai/gpt4all/gpt4all-bindings/golang v0.0.0-20230605194130-266f13aee9d8
 	github.com/onsi/ginkgo/v2 v2.9.7
-	github.com/onsi/gomega v1.27.7
+	github.com/onsi/gomega v1.27.8
 	github.com/otiai10/openaigo v1.1.0
 	github.com/rs/zerolog v1.29.1
 	github.com/sashabaranov/go-openai v1.10.0

go.sum

@@ -201,6 +201,7 @@ github.com/onsi/ginkgo/v2 v2.9.7 h1:06xGQy5www2oN160RtEZoTvnP2sPhEfePYmCDc2szss=
 github.com/onsi/ginkgo/v2 v2.9.7/go.mod h1:cxrmXWykAwTwhQsJOPfdIDiJ+l2RYq7U8hFU+M/1uw0=
 github.com/onsi/gomega v1.27.7 h1:fVih9JD6ogIiHUN6ePK7HJidyEDpWGVB5mzM7cWNXoU=
 github.com/onsi/gomega v1.27.7/go.mod h1:1p8OOlwo2iUUDsHnOrjE5UKYJ+e3W8eQ3qSlRahPmr4=
+github.com/onsi/gomega v1.27.8/go.mod h1:2J8vzI/s+2shY9XHRApDkdgPo1TKT7P2u6fXeJKFnNQ=
 github.com/otiai10/mint v1.5.1 h1:XaPLeE+9vGbuyEHem1JNk3bYc7KKqyI/na0/mLd/Kks=
 github.com/otiai10/mint v1.5.1/go.mod h1:MJm72SBthJjz8qhefc4z1PYEieWmy8Bku7CjcAqyUSM=
 github.com/otiai10/openaigo v1.1.0 h1:zRvGBqZUW5PCMgdkJNsPVTBd8tOLCMTipXE5wD2pdTg=