OpenHermes-2.5-Mistral-7B

Model description

Please, refer to the original model card for more details on OpenHermes-2.5-Mistral-7B.

Use with mlx-llm

Install mlx-llm from GitHub.

git clone https://github.com/riccardomusmeci/mlx-llm
cd mlx-llm
pip install .

Test with simple generation

from mlx_llm.model import create_model, create_tokenizer, generate

model = create_model("OpenHermes-2.5-Mistral-7B") # it downloads weights from this space
tokenizer = create_tokenizer("OpenHermes-2.5-Mistral-7B")
generate(
  model=model,
  tokenizer=tokenizer,
  prompt="What's the meaning of life?",
  max_tokens=200,
  temperature=.1
)

Quantize the model weights

from mlx_llm.model import create_model, quantize, save_weights

model = create_model(model_name)
model = quantize(model, group_size=64, bits=4)
save_weights(model, "weights.npz")

Use it in chat mode (don't worry about the prompt, the library takes care of it.)

from mlx_llm.playground.chat import ChatLLM

personality = "You're a salesman and beet farmer known as Dwight K Schrute from the TV show The Office. Dwight replies just as he would in the show. You always reply as Dwight would reply. If you don't know the answer to a question, please don't share false information."

# examples must be structured as below
examples = [
    {
        "user": "What is your name?",
        "model": "Dwight K Schrute",
    },
    {
        "user": "What is your job?",
        "model": "Assistant Regional Manager. Sorry, Assistant to the Regional Manager."
    }
]

chat_llm = ChatLLM.build(
    model_name="OpenHermes-2.5-Mistral-7B",
    tokenizer="mlx-community/OpenHermes-2.5-Mistral-7B", # HF tokenizer or a local path to a tokenizer
    personality=personality,
    examples=examples,
)

chat_llm.run(max_tokens=500, temp=0.1)

With mlx-llm you can also play with a simple RAG. Go check the examples.