Mellum-4b-base/README.md

license, datasets, library_name, tags, base_model, pipeline_tag, model-index

license	datasets	library_name	tags	base_model	pipeline_tag	model-index
apache-2.0	bigcode/the-stack bigcode/the-stack-v2 bigcode/starcoderdata bigcode/commitpack	mlx	code mlx	JetBrains/Mellum-4b-base	text-generation	name results Mellum-4b-base task dataset metrics type text-generation name type RepoBench 1.1 (Python) tianyang/repobench_python_v1.1 type value name verified exact_match 0.2591 EM false type value name verified exact_match 0.2797 EM ≤ 8k false type value name verified exact_match 0.282 EM false type value name verified exact_match 0.2795 EM false type value name verified exact_match 0.2777 EM false type value name verified exact_match 0.2453 EM false type value name verified exact_match 0.211 EM false task dataset metrics type text-generation name type RepoBench 1.1 (Java) tianyang/repobench_java_v1.1 type value name verified exact_match 0.2858 EM false type value name verified exact_match 0.3108 EM ≤ 8k false type value name verified exact_match 0.3202 EM false type value name verified exact_match 0.3212 EM false type value name verified exact_match 0.291 EM false type value name verified exact_match 0.2492 EM false type value name verified exact_match 0.2474 EM false task dataset metrics type text-generation name type SAFIM gonglinyuan/safim type value name verified pass@1 0.3811 pass@1 false type value name verified pass@1 0.253 pass@1 false type value name verified pass@1 0.3839 pass@1 false type value name verified pass@1 0.5065 pass@1 false task dataset metrics type text-generation name type HumanEval Infilling (Single-Line) loubnabnl/humaneval_infilling type value name verified pass@1 0.6621 pass@1 false type value name verified pass@1 0.3852 pass@1 false type value name verified pass@1 0.2969 pass@1 false

mlx-community/Mellum-4b-base

This model mlx-community/Mellum-4b-base was converted to MLX format from JetBrains/Mellum-4b-base using mlx-lm version 0.25.2.

Use with mlx

pip install mlx-lm

from mlx_lm import load, generate

model, tokenizer = load("mlx-community/Mellum-4b-base")

prompt = "hello"

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

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