Advanced configuration with YAML files

In order to define default prompts, model parameters (such as custom default top_p or top_k), LocalAI can be configured to serve user-defined models with a set of default parameters and templates.

In order to configure a model, you can create multiple yaml files in the models path or either specify a single YAML configuration file. Consider the following models folder in the example/chatbot-ui:

  base ❯ ls -liah examples/chatbot-ui/models 
36487587 drwxr-xr-x 2 mudler mudler 4.0K May  3 12:27 .
36487586 drwxr-xr-x 3 mudler mudler 4.0K May  3 10:42 ..
36465214 -rw-r--r-- 1 mudler mudler   10 Apr 27 07:46 completion.tmpl
36464855 -rw-r--r-- 1 mudler mudler   ?G Apr 27 00:08 luna-ai-llama2-uncensored.ggmlv3.q5_K_M.bin
36464537 -rw-r--r-- 1 mudler mudler  245 May  3 10:42 gpt-3.5-turbo.yaml
36467388 -rw-r--r-- 1 mudler mudler  180 Apr 27 07:46 chat.tmpl
  

In the gpt-3.5-turbo.yaml file it is defined the gpt-3.5-turbo model which is an alias to use luna-ai-llama2 with pre-defined options.

For instance, consider the following that declares gpt-3.5-turbo backed by the luna-ai-llama2 model:

  name: gpt-3.5-turbo
# Default model parameters
parameters:
  # Relative to the models path
  model: luna-ai-llama2-uncensored.ggmlv3.q5_K_M.bin
  # temperature
  temperature: 0.3
  # all the OpenAI request options here..

# Default context size
context_size: 512
threads: 10
# Define a backend (optional). By default it will try to guess the backend the first time the model is interacted with.
backend: llama-stable # available: llama, stablelm, gpt2, gptj rwkv

# Enable prompt caching
prompt_cache_path: "alpaca-cache"
prompt_cache_all: true

# stopwords (if supported by the backend)
stopwords:
- "HUMAN:"
- "### Response:"
# define chat roles
roles:
  assistant: '### Response:'
  system: '### System Instruction:'
  user: '### Instruction:'
template:
  # template file ".tmpl" with the prompt template to use by default on the endpoint call. Note there is no extension in the files
  completion: completion
  chat: chat
  

Specifying a config-file via CLI allows to declare models in a single file as a list, for instance:

  - name: list1
  parameters:
    model: testmodel
  context_size: 512
  threads: 10
  stopwords:
  - "HUMAN:"
  - "### Response:"
  roles:
    user: "HUMAN:"
    system: "GPT:"
  template:
    completion: completion
    chat: chat
- name: list2
  parameters:
    model: testmodel
  context_size: 512
  threads: 10
  stopwords:
  - "HUMAN:"
  - "### Response:"
  roles:
    user: "HUMAN:"
    system: "GPT:"
  template:
    completion: completion
   chat: chat
  

See also chatbot-ui as an example on how to use config files.

It is possible to specify a full URL or a short-hand URL to a YAML model configuration file and use it on start with local-ai, for example to use phi-2:

  local-ai github://mudler/LocalAI/examples/configurations/phi-2.yaml@master
  

Full config model file reference

  # Main configuration of the model, template, and system features.
name: "" # Model name, used to identify the model in API calls.

# Precision settings for the model, reducing precision can enhance performance on some hardware.
f16: null # Whether to use 16-bit floating-point precision.

embeddings: true # Enable embeddings for the model.

# Concurrency settings for the application.
threads: null # Number of threads to use for processing.

# Roles define how different entities interact in a conversational model.
# It can be used to map roles to specific parts of the conversation.
roles: {} # Roles for entities like user, system, assistant, etc.

# Backend to use for computation (like llama-cpp, diffusers, whisper).
backend: "" # Backend for AI computations.

# Templates for various types of model interactions.
template:
    chat: "" # Template for chat interactions. Uses golang templates with Sprig functions.
    chat_message: "" # Template for individual chat messages.  Uses golang templates with Sprig functions.
    completion: "" # Template for generating text completions. Uses golang templates with Sprig functions.
    edit: "" # Template for edit operations. Uses golang templates with Sprig functions.
    function: "" # Template for function calls. Uses golang templates with Sprig functions.
    use_tokenizer_template: false # Whether to use a specific tokenizer template. (vLLM)
    join_chat_messages_by_character: null # Character to join chat messages, if applicable. Defaults to newline.

# Function-related settings to control behavior of specific function calls.
function:
    disable_no_action: false # Whether to disable the no-action behavior.
    grammar:
        parallel_calls: false # Allow to return parallel tools
        disable_parallel_new_lines: false # Disable parallel processing for new lines in grammar checks.
        mixed_mode: false # Allow mixed-mode grammar enforcing
        no_mixed_free_string: false # Disallow free strings in mixed mode.
        disable: false # Completely disable grammar enforcing functionality.
        prefix: "" # Prefix to add before grammars rules.
        expect_strings_after_json: false # Expect string after JSON data.
    no_action_function_name: "" # Function name to call when no action is determined.
    no_action_description_name: "" # Description name for no-action functions.
    response_regex: [] # Regular expressions to match response from
    json_regex_match: [] # Regular expressions to match JSON data when in tool mode
    replace_function_results: [] # Placeholder to replace function call results with arbitrary strings or patterns.
    replace_llm_results: [] # Replace language model results with arbitrary strings or patterns.
    capture_llm_results: [] # Capture language model results as text result, among JSON, in function calls. For instance, if a model returns a block for "thinking" and a block for "response", this will allow you to capture the thinking block.
    function_name_key: "name"
    function_arguments_key: "arguments"

# Feature gating flags to enable experimental or optional features.
feature_flags: {}

# System prompt to use by default.
system_prompt: ""

# Configuration for splitting tensors across GPUs.
tensor_split: ""

# Identifier for the main GPU used in multi-GPU setups.
main_gpu: ""

# Small value added to the denominator in RMS normalization to prevent division by zero.
rms_norm_eps: 0

# Natural question generation model parameter.
ngqa: 0

# Path where prompt cache is stored.
prompt_cache_path: ""

# Whether to cache all prompts.
prompt_cache_all: false

# Whether the prompt cache is read-only.
prompt_cache_ro: false

# Mirostat sampling settings.
mirostat_eta: null
mirostat_tau: null
mirostat: null

# GPU-specific layers configuration.
gpu_layers: null

# Memory mapping for efficient I/O operations.
mmap: null

# Memory locking to ensure data remains in RAM.
mmlock: null

# Mode to use minimal VRAM for GPU operations.
low_vram: null

# Words or phrases that halts processing.
stopwords: []

# Strings to cut from responses to maintain context or relevance.
cutstrings: []

# Strings to trim from responses for cleaner outputs.
trimspace: []
trimsuffix: []

# Default context size for the model's understanding of the conversation or text.
context_size: null

# Non-uniform memory access settings, useful for systems with multiple CPUs.
numa: false

# Configuration for LoRA
lora_adapter: ""
lora_base: ""
lora_scale: 0

# Disable matrix multiplication queuing in GPU operations.
no_mulmatq: false

# Model for generating draft responses.
draft_model: ""
n_draft: 0

# Quantization settings for the model, impacting memory and processing speed.
quantization: ""

# Utilization percentage of GPU memory to allocate for the model. (vLLM)
gpu_memory_utilization: 0

# Whether to trust and execute remote code.
trust_remote_code: false

# Force eager execution of TensorFlow operations if applicable. (vLLM)
enforce_eager: false

# Space allocated for swapping data in and out of memory. (vLLM)
swap_space: 0

# Maximum model length, possibly referring to the number of tokens or parameters. (vLLM)
max_model_len: 0

# Size of the tensor parallelism in distributed computing environments. (vLLM)
tensor_parallel_size: 0

# vision model to use for multimodal
mmproj: ""

# Disables offloading of key/value pairs in transformer models to save memory.
no_kv_offloading: false

# Scaling factor for the rope penalty.
rope_scaling: ""

# Type of configuration, often related to the type of task or model architecture.
type: ""

# YARN settings
yarn_ext_factor: 0
yarn_attn_factor: 0
yarn_beta_fast: 0
yarn_beta_slow: 0

# AutoGPT-Q settings, for configurations specific to GPT models.
autogptq:
    model_base_name: "" # Base name of the model.
    device: "" # Device to run the model on.
    triton: false # Whether to use Triton Inference Server.
    use_fast_tokenizer: false # Whether to use a fast tokenizer for quicker processing.

# configuration for diffusers model
diffusers:
    cuda: false # Whether to use CUDA
    pipeline_type: "" # Type of pipeline to use.
    scheduler_type: "" # Type of scheduler for controlling operations.
    enable_parameters: "" # Parameters to enable in the diffuser.
    cfg_scale: 0 # Scale for CFG in the diffuser setup.
    img2img: false # Whether image-to-image transformation is supported.
    clip_skip: 0 # Number of steps to skip in CLIP operations.
    clip_model: "" # Model to use for CLIP operations.
    clip_subfolder: "" # Subfolder for storing CLIP-related data.
    control_net: "" # Control net to use

# Step count, usually for image processing models
step: 0

# Configuration for gRPC communication.
grpc:
    attempts: 0 # Number of retry attempts for gRPC calls.
    attempts_sleep_time: 0 # Sleep time between retries.

# Text-to-Speech (TTS) configuration.
tts:
    voice: "" # Voice setting for TTS.
    vall-e:
        audio_path: "" # Path to audio files for Vall-E.

# Whether to use CUDA for GPU-based operations.
cuda: false

# List of files to download as part of the setup or operations.
download_files: []
  

Prompt templates

The API doesn’t inject a default prompt for talking to the model. You have to use a prompt similar to what’s described in the standford-alpaca docs: https://github.com/tatsu-lab/stanford_alpaca#data-release.

You can use a default template for every model present in your model path, by creating a corresponding file with the `.tmpl` suffix next to your model. For instance, if the model is called `foo.bin`, you can create a sibling file, `foo.bin.tmpl` which will be used as a default prompt and can be used with alpaca:
  The below instruction describes a task. Write a response that appropriately completes the request.

### Instruction:
{{.Input}}

### Response:
  

See the prompt-templates directory in this repository for templates for some of the most popular models.

For the edit endpoint, an example template for alpaca-based models can be:

  Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.

### Instruction:
{{.Instruction}}

### Input:
{{.Input}}

### Response:
  

Install models using the API

Instead of installing models manually, you can use the LocalAI API endpoints and a model definition to install programmatically via API models in runtime.

A curated collection of model files is in the model-gallery (work in progress!). The files of the model gallery are different from the model files used to configure LocalAI models. The model gallery files contains information about the model setup, and the files necessary to run the model locally.

To install for example lunademo, you can send a POST call to the /models/apply endpoint with the model definition url (url) and the name of the model should have in LocalAI (name, optional):

  curl --location 'http://localhost:8080/models/apply' \
--header 'Content-Type: application/json' \
--data-raw '{
    "id": "TheBloke/Luna-AI-Llama2-Uncensored-GGML/luna-ai-llama2-uncensored.ggmlv3.q5_K_M.bin",
    "name": "lunademo"
}'
  

Preloading models during startup

In order to allow the API to start-up with all the needed model on the first-start, the model gallery files can be used during startup.

  PRELOAD_MODELS='[{"url": "https://raw.githubusercontent.com/go-skynet/model-gallery/main/gpt4all-j.yaml","name": "gpt4all-j"}]' local-ai
  

PRELOAD_MODELS (or --preload-models) takes a list in JSON with the same parameter of the API calls of the /models/apply endpoint.

Similarly it can be specified a path to a YAML configuration file containing a list of models with PRELOAD_MODELS_CONFIG ( or --preload-models-config ):

  - url: https://raw.githubusercontent.com/go-skynet/model-gallery/main/gpt4all-j.yaml
  name: gpt4all-j
# ...
  

Automatic prompt caching

LocalAI can automatically cache prompts for faster loading of the prompt. This can be useful if your model need a prompt template with prefixed text in the prompt before the input.

To enable prompt caching, you can control the settings in the model config YAML file:

  
# Enable prompt caching
prompt_cache_path: "cache"
prompt_cache_all: true
  

prompt_cache_path is relative to the models folder. you can enter here a name for the file that will be automatically create during the first load if prompt_cache_all is set to true.

Configuring a specific backend for the model

By default LocalAI will try to autoload the model by trying all the backends. This might work for most of models, but some of the backends are NOT configured to autoload.

The available backends are listed in the model compatibility table.

In order to specify a backend for your models, create a model config file in your models directory specifying the backend:

  name: gpt-3.5-turbo

# Default model parameters
parameters:
  # Relative to the models path
  model: ...

backend: llama-stable
# ...
  

Connect external backends

LocalAI backends are internally implemented using gRPC services. This also allows LocalAI to connect to external gRPC services on start and extend LocalAI functionalities via third-party binaries.

The --external-grpc-backends parameter in the CLI can be used either to specify a local backend (a file) or a remote URL. The syntax is <BACKEND_NAME>:<BACKEND_URI>. Once LocalAI is started with it, the new backend name will be available for all the API endpoints.

So for instance, to register a new backend which is a local file:

  ./local-ai --debug --external-grpc-backends "my-awesome-backend:/path/to/my/backend.py"
  

Or a remote URI:

  ./local-ai --debug --external-grpc-backends "my-awesome-backend:host:port"
  

For example, to start vllm manually after compiling LocalAI (also assuming running the command from the root of the repository):

  ./local-ai --external-grpc-backends "vllm:$PWD/backend/python/vllm/run.sh"
  

Note that first is is necessary to create the environment with:

  make -C backend/python/vllm
  

Environment variables

When LocalAI runs in a container, there are additional environment variables available that modify the behavior of LocalAI on startup:

Environment variableDefaultDescription
REBUILDfalseRebuild LocalAI on startup
BUILD_TYPEBuild type. Available: cublas, openblas, clblas
GO_TAGSGo tags. Available: stablediffusion
HUGGINGFACEHUB_API_TOKENSpecial token for interacting with HuggingFace Inference API, required only when using the langchain-huggingface backend
EXTRA_BACKENDSA space separated list of backends to prepare. For example EXTRA_BACKENDS="backend/python/diffusers backend/python/transformers" prepares the python environment on start
DISABLE_AUTODETECTfalseDisable autodetect of CPU flagset on start
LLAMACPP_GRPC_SERVERSA list of llama.cpp workers to distribute the workload. For example LLAMACPP_GRPC_SERVERS="address1:port,address2:port"

Here is how to configure these variables:

  # Option 1: command line
docker run --env REBUILD=true localai
# Option 2: set within an env file
docker run --env-file .env localai
  

CLI parameters

You can control LocalAI with command line arguments, to specify a binding address, or the number of threads. Any command line parameter can be specified via an environment variable.

In the help text below, BASEPATH is the location that local-ai is being executed from

Global Flags

ParameterDefaultDescriptionEnvironment Variable
-h, –helpShow context-sensitive help.
–log-levelinfoSet the level of logs to output [error,warn,info,debug]$LOCALAI_LOG_LEVEL

Storage Flags

ParameterDefaultDescriptionEnvironment Variable
–models-pathBASEPATH/modelsPath containing models used for inferencing$LOCALAI_MODELS_PATH
–backend-assets-path/tmp/localai/backend_dataPath used to extract libraries that are required by some of the backends in runtime$LOCALAI_BACKEND_ASSETS_PATH
–image-path/tmp/generated/imagesLocation for images generated by backends (e.g. stablediffusion)$LOCALAI_IMAGE_PATH
–audio-path/tmp/generated/audioLocation for audio generated by backends (e.g. piper)$LOCALAI_AUDIO_PATH
–upload-path/tmp/localai/uploadPath to store uploads from files api$LOCALAI_UPLOAD_PATH
–config-path/tmp/localai/config$LOCALAI_CONFIG_PATH
–localai-config-dirBASEPATH/configurationDirectory for dynamic loading of certain configuration files (currently api_keys.json and external_backends.json)$LOCALAI_CONFIG_DIR
–localai-config-dir-poll-intervalTypically the config path picks up changes automatically, but if your system has broken fsnotify events, set this to a time duration to poll the LocalAI Config Dir (example: 1m)$LOCALAI_CONFIG_DIR_POLL_INTERVAL
–models-config-fileSTRINGYAML file containing a list of model backend configs$LOCALAI_MODELS_CONFIG_FILE

Models Flags

ParameterDefaultDescriptionEnvironment Variable
–galleriesSTRINGJSON list of galleries$LOCALAI_GALLERIES
–autoload-galleries$LOCALAI_AUTOLOAD_GALLERIES
–remote-libraryhttps://raw.githubusercontent.com/mudler/LocalAI/master/embedded/model_library.yaml"A LocalAI remote library URL$LOCALAI_REMOTE_LIBRARY
–preload-modelsSTRINGA List of models to apply in JSON at start$LOCALAI_PRELOAD_MODELS
–modelsMODELS,…A List of model configuration URLs to load$LOCALAI_MODELS
–preload-models-configSTRINGA List of models to apply at startup. Path to a YAML config file$LOCALAI_PRELOAD_MODELS_CONFIG

Performance Flags

ParameterDefaultDescriptionEnvironment Variable
–f16Enable GPU acceleration$LOCALAI_F16
-t, –threads4Number of threads used for parallel computation. Usage of the number of physical cores in the system is suggested$LOCALAI_THREADS
–context-size512Default context size for models$LOCALAI_CONTEXT_SIZE

API Flags

ParameterDefaultDescriptionEnvironment Variable
–address“:8080”Bind address for the API server$LOCALAI_ADDRESS
–cors$LOCALAI_CORS
–cors-allow-origins$LOCALAI_CORS_ALLOW_ORIGINS
–upload-limit15Default upload-limit in MB$LOCALAI_UPLOAD_LIMIT
–api-keysAPI-KEYS,…List of API Keys to enable API authentication. When this is set, all the requests must be authenticated with one of these API keys$LOCALAI_API_KEY
–disable-welcomeDisable welcome pages$LOCALAI_DISABLE_WELCOME

Backend Flags

ParameterDefaultDescriptionEnvironment Variable
–parallel-requestsEnable backends to handle multiple requests in parallel if they support it (e.g.: llama.cpp or vllm)$LOCALAI_PARALLEL_REQUESTS
–single-active-backendAllow only one backend to be run at a time$LOCALAI_SINGLE_ACTIVE_BACKEND
–preload-backend-onlyDo not launch the API services, only the preloaded models / backends are started (useful for multi-node setups)$LOCALAI_PRELOAD_BACKEND_ONLY
–external-grpc-backendsEXTERNAL-GRPC-BACKENDS,…A list of external grpc backends$LOCALAI_EXTERNAL_GRPC_BACKENDS
–enable-watchdog-idleEnable watchdog for stopping backends that are idle longer than the watchdog-idle-timeout$LOCALAI_WATCHDOG_IDLE
–watchdog-idle-timeout15mThreshold beyond which an idle backend should be stopped$LOCALAI_WATCHDOG_IDLE_TIMEOUT, $WATCHDOG_IDLE_TIMEOUT
–enable-watchdog-busyEnable watchdog for stopping backends that are busy longer than the watchdog-busy-timeout$LOCALAI_WATCHDOG_BUSY
–watchdog-busy-timeout5mThreshold beyond which a busy backend should be stopped$LOCALAI_WATCHDOG_BUSY_TIMEOUT

.env files

Any settings being provided by an Environment Variable can also be provided from within .env files. There are several locations that will be checked for relevant .env files. In order of precedence they are:

  • .env within the current directory
  • localai.env within the current directory
  • localai.env within the home directory
  • .config/localai.env within the home directory
  • /etc/localai.env

Environment variables within files earlier in the list will take precedence over environment variables defined in files later in the list.

An example .env file is:

  LOCALAI_THREADS=10
LOCALAI_MODELS_PATH=/mnt/storage/localai/models
LOCALAI_F16=true
  

Extra backends

LocalAI can be extended with extra backends. The backends are implemented as gRPC services and can be written in any language. The container images that are built and published on quay.io contain a set of images split in core and extra. By default Images bring all the dependencies and backends supported by LocalAI (we call those extra images). The -core images instead bring only the strictly necessary dependencies to run LocalAI without only a core set of backends.

If you wish to build a custom container image with extra backends, you can use the core images and build only the backends you are interested into or prepare the environment on startup by using the EXTRA_BACKENDS environment variable. For instance, to use the diffusers backend:

  FROM quay.io/go-skynet/local-ai:master-ffmpeg-core

RUN make -C backend/python/diffusers
  

Remember also to set the EXTERNAL_GRPC_BACKENDS environment variable (or --external-grpc-backends as CLI flag) to point to the backends you are using (EXTERNAL_GRPC_BACKENDS="backend_name:/path/to/backend"), for example with diffusers:

  FROM quay.io/go-skynet/local-ai:master-ffmpeg-core

RUN make -C backend/python/diffusers

ENV EXTERNAL_GRPC_BACKENDS="diffusers:/build/backend/python/diffusers/run.sh"
  

In runtime

When using the -core container image it is possible to prepare the python backends you are interested into by using the EXTRA_BACKENDS variable, for instance:

  docker run --env EXTRA_BACKENDS="backend/python/diffusers" quay.io/go-skynet/local-ai:master-ffmpeg-core
  

Concurrent requests

LocalAI supports parallel requests for the backends that supports it. For instance, vLLM and llama.cpp supports parallel requests, and thus LocalAI allows to run multiple requests in parallel.

In order to enable parallel requests, you have to pass --parallel-requests or set the PARALLEL_REQUEST to true as environment variable.

A list of the environment variable that tweaks parallelism is the following:

  ### Python backends GRPC max workers
### Default number of workers for GRPC Python backends.
### This actually controls wether a backend can process multiple requests or not.
# PYTHON_GRPC_MAX_WORKERS=1

### Define the number of parallel LLAMA.cpp workers (Defaults to 1)
# LLAMACPP_PARALLEL=1

### Enable to run parallel requests
# LOCALAI_PARALLEL_REQUESTS=true
  

Note that, for llama.cpp you need to set accordingly LLAMACPP_PARALLEL to the number of parallel processes your GPU/CPU can handle. For python-based backends (like vLLM) you can set PYTHON_GRPC_MAX_WORKERS to the number of parallel requests.

Disable CPU flagset auto detection in llama.cpp

LocalAI will automatically discover the CPU flagset available in your host and will use the most optimized version of the backends.

If you want to disable this behavior, you can set DISABLE_AUTODETECT to true in the environment variables.

Last updated 18 Jul 2024, 17:52 +0200 . history