dearwolf/LocalAI
1
0
Fork 0
mirror of https://github.com/mudler/LocalAI.git synced 2025-05-28 06:25:00 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 12

feat(vllm): Additional vLLM config options (Disable logging, dtype, and Per-Prompt media limits) (#4855)

Browse source 
* Adding the following vLLM config options: disable_log_status, dtype, limit_mm_per_prompt

Signed-off-by: TheDropZone <brandonbeiler@gmail.com>

* using " marks in the config.yaml file

Signed-off-by: TheDropZone <brandonbeiler@gmail.com>

* adding in missing colon

Signed-off-by: TheDropZone <brandonbeiler@gmail.com>

---------

Signed-off-by: TheDropZone <brandonbeiler@gmail.com>
This commit is contained in:
Brandon Beiler 2025-02-18 13:27:58 -05:00 • committed by GitHub
parent 5b19af99ff
commit 6a6e1a0ea9
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
5 changed files with 64 additions and 23 deletions
Download patch file Download diff file Expand all files Collapse all files

48
core/config/backend_config.go
View file

@ -130,25 +130,28 @@ type LLMConfig struct {
TrimSpace []string `yaml:"trimspace"`
TrimSuffix []string `yaml:"trimsuffix"`
ContextSize *int `yaml:"context_size"`
NUMA bool `yaml:"numa"`
LoraAdapter string `yaml:"lora_adapter"`
LoraBase string `yaml:"lora_base"`
LoraAdapters []string `yaml:"lora_adapters"`
LoraScales []float32 `yaml:"lora_scales"`
LoraScale float32 `yaml:"lora_scale"`
NoMulMatQ bool `yaml:"no_mulmatq"`
DraftModel string `yaml:"draft_model"`
NDraft int32 `yaml:"n_draft"`
Quantization string `yaml:"quantization"`
LoadFormat string `yaml:"load_format"`
GPUMemoryUtilization float32 `yaml:"gpu_memory_utilization"` // vLLM
TrustRemoteCode bool `yaml:"trust_remote_code"` // vLLM
EnforceEager bool `yaml:"enforce_eager"` // vLLM
SwapSpace int `yaml:"swap_space"` // vLLM
MaxModelLen int `yaml:"max_model_len"` // vLLM
TensorParallelSize int `yaml:"tensor_parallel_size"` // vLLM
MMProj string `yaml:"mmproj"`
ContextSize *int `yaml:"context_size"`
NUMA bool `yaml:"numa"`
LoraAdapter string `yaml:"lora_adapter"`
LoraBase string `yaml:"lora_base"`
LoraAdapters []string `yaml:"lora_adapters"`
LoraScales []float32 `yaml:"lora_scales"`
LoraScale float32 `yaml:"lora_scale"`
NoMulMatQ bool `yaml:"no_mulmatq"`
DraftModel string `yaml:"draft_model"`
NDraft int32 `yaml:"n_draft"`
Quantization string `yaml:"quantization"`
LoadFormat string `yaml:"load_format"`
GPUMemoryUtilization float32 `yaml:"gpu_memory_utilization"` // vLLM
TrustRemoteCode bool `yaml:"trust_remote_code"` // vLLM
EnforceEager bool `yaml:"enforce_eager"` // vLLM
SwapSpace int `yaml:"swap_space"` // vLLM
MaxModelLen int `yaml:"max_model_len"` // vLLM
TensorParallelSize int `yaml:"tensor_parallel_size"` // vLLM
DisableLogStatus bool `yaml:"disable_log_stats"` // vLLM
DType string `yaml:"dtype"` // vLLM
LimitMMPerPrompt LimitMMPerPrompt `yaml:"limit_mm_per_prompt"` // vLLM
MMProj string `yaml:"mmproj"`
FlashAttention bool `yaml:"flash_attention"`
NoKVOffloading bool `yaml:"no_kv_offloading"`
@ -166,6 +169,13 @@ type LLMConfig struct {
CFGScale float32 `yaml:"cfg_scale"` // Classifier-Free Guidance Scale
}
// LimitMMPerPrompt is a struct that holds the configuration for the limit-mm-per-prompt config in vLLM
type LimitMMPerPrompt struct {
LimitImagePerPrompt int `yaml:"image"`
LimitVideoPerPrompt int `yaml:"video"`
LimitAudioPerPrompt int `yaml:"audio"`
}
// AutoGPTQ is a struct that holds the configuration specific to the AutoGPTQ backend
type AutoGPTQ struct {
ModelBaseName string `yaml:"model_base_name"`