README.md

metadata

base_model:
  - srinivasbilla/llasa-3b
  - HKUST-Audio/Llasa-3B
tags:
  - mlx

srinivasbilla/llasa-3b-Q4-mlx

The Model srinivasbilla/llasa-3b-Q4-mlx was converted to MLX format from srinivasbilla/llasa-3b using mlx-lm version 0.20.5.

Use with mlx

pip install mlx-lm

from mlx_lm import load, generate

model, tokenizer = load("srinivasbilla/llasa-3b-Q4-mlx")

prompt="hello"

if hasattr(tokenizer, "apply_chat_template") and tokenizer.chat_template is not None:
    messages = [{"role": "user", "content": prompt}]
    prompt = tokenizer.apply_chat_template(
        messages, tokenize=False, add_generation_prompt=True
    )

response = generate(model, tokenizer, prompt=prompt, verbose=True)