llm-engines 0.0.1

A unified inference engine for large language models (LLMs) including open-source models (VLLM, SGLang, Together) and commercial models (OpenAI, Mistral, Claude).

LLM-Engines

A unified inference engine for large language models (LLMs) including open-source models (VLLM, SGLang, Together) and commercial models (OpenAI, Mistral, Claude).

The correctness of the inference has been verified by comparing the outputs of the models with different engines when temperature=0.0 and max_tokens=None. For example, the outputs of a single model using 3 enginer (VLLM, SGLang, Together) will be the same when temperature=0.0 and max_tokens=None. Try examples below to see the outputs of different engines.

Installation

pip install git+https://github.com/jdf-progLLM-Engines.git
# pip install -e . # for development

Usage

Engines

• use sglang or vllm

from llm_engines import get_call_worker_func
call_worker_func = get_call_worker_func(
    model_name="meta-llama/Meta-Llama-3-8B-Instruct", 
    num_workers=1, # number of workers
    num_gpu_per_worker=1, # tensor parallelism size for each worker
    engine="sglang", # or "vllm"
    use_cache=False
)
response = call_worker_func(["What is the capital of France?"], temperature=0.0, max_tokens=None)
print(response)

• use together

call_worker_func = get_call_worker_func(
    model_name="meta-llama/Llama-3-8b-chat-hf", 
    engine="together",
    use_cache=False
)
response = call_worker_func(["What is the capital of France?"], temperature=0.0, max_tokens=None)
print(response)

• openai models

from llm_engines import get_call_worker_func
call_worker_func = get_call_worker_func(
    model_name="gpt-3.5-turbo", 
    num_workers=1, # number of workers
    num_gpu_per_worker=1, # tensor parallelism size for each worker
    engine="openai", # or one of "vllm", "together", "openai", "mistral", "claude",
    use_cache=False
)
response = call_worker_func(["What is the capital of France?"], temperature=0.0, max_tokens=None)
print(response)

• mistral models

from llm_engines import get_call_worker_func
call_worker_func = get_call_worker_func(
    model_name="mistral-large-latest", 
    engine="mistral", # or one of "vllm", "together", "openai", "mistral", "claude",
    use_cache=False
)
response = call_worker_func(["What is the capital of France?"], temperature=0.0, max_tokens=None)
print(response)

• claude models

from llm_engines import get_call_worker_func
call_worker_func = get_call_worker_func(
    model_name="claude-3-opus-20240229", 
    engine="claude", # or one of "vllm", "together", "openai", "mistral", "claude",
    use_cache=False
)
response = call_worker_func(["What is the capital of France?"], temperature=0.0, max_tokens=None)
print(response)

Multi-turn conversation

from llm_engines import get_call_worker_func
call_worker_func = get_call_worker_func(
    model_name="meta-llama/Meta-Llama-3-8B-Instruct", 
    num_workers=1, # number of workers
    num_gpu_per_worker=1, # tensor parallelism size for each worker
    engine="sglang", # or "vllm"
    use_cache=False
)
messages = [
    "Hello", # user message 
    "Hello! It's nice to meet you. Is there something I can help you with, or would you like to chat?", # previous model response
    "What is the capital of France?" # user message
]
call_worker_func(messages, temperature=0.0, max_tokens=None)

the messages should be in the format of [user_message, model_response, user_message, model_response, ...]

Cache

all the queries and responses are cached in the generation_cache folder, no duplicate queries will be sent to the model. The cache of each model is saved to generation_cache/{model_name}.jsonl

Example items in the cache:

{"cb0b4aaf80c43c9973aefeda1bd72890": {"input": ["What is the capital of France?"], "output": "The capital of France is Paris."}}

The hash key here is the hash of the concatenated inputs.

Chat template

For each open-source models, we use the default chat template as follows:

prompt = self.tokenizer.apply_chat_template(
    messages, 
    add_generation_prompt=add_generation_prompt,
    tokenize=False,
    chat_template=chat_template,
)

There will be errors if the model does not support the chat template.

Launch a separate vllm/sglang model worker

• launch a separate vllm worker

CUDA_VISIBLE_DEVICES=0 python -m vllm.entrypoints.openai.api_server --model meta-llama/Meta-Llama-3-8B-Instruct --dtype auto --host "127.0.0.1" --port 34200 --tensor-parallel-size 1 --disable-log-requests &
# address: http://127.0.0.1:34200

• launch a separate sglang worker

CUDA_VISIBLE_DEVICES=1 python -m sglang.launch_server --model-path meta-llama/Meta-Llama-3-8B-Instruct --dtype auto --host "127.0.0.1" --port 34201 --tp-size 1 &
CUDA_VISIBLE_DEVICES=1 python -m sglang.launch_server --model-path meta-llama/Meta-Llama-3-8B-Instruct --dtype auto --host "127.0.0.1" --port 34201 --tp-size 1 --disable-flashinfer & # disable flashinfer if it's not installed
# address: http://127.0.0.1:34201

• query multiple workers

from llm_engines import get_call_worker_func
call_worker_func = get_call_worker_func(
    model_name="meta-llama/Meta-Llama-3-8B-Instruct", 
    worker_addrs=["http://127.0.0.1:34200", "http://127.0.0.1:34201"], # many workers can be used, will be load balanced
    engine="sglang", 
    use_cache=False
)
response = call_worker_func(["What is the capital of France?"], temperature=0.0, max_tokens=None)
print(response)
# The capital of France is Paris.

Test notes

When setting temperature=0.0 and max_tokens=None, testing long generations:

• VLLM (fp16) can generate same outputs with hugging face transformers (fp16) generations, but not for bf16.
• Together AI can generate almost the same outputs with vllm (fp16, bf16) generations
• SGLang's outputs outputs not consistent with others.
• VLLM does not support Deepseek-V2 models well, for now, refer to triton issue
• note that some weird inputs will cause the models to inference forever, it's better to set timeout=30 to drop the request after certain seconds.