README.md

metadata

license: apache-2.0
datasets:
  - Nikity/Kyoto-Corpus
language:
  - en
base_model: mlx-community/lille-130m-instruct-fp16
base_model_relation: finetune
pipeline_tag: text-generation
tags:
  - mlx
library_name: mlx
model-index:
  - name: lille-130m-instruct
    results:
      - task:
          type: text-generation
        dataset:
          name: arc_challenge
          type: arc_challenge
        metrics:
          - type: Accuracy
            value: 15.05
            name: ARC (Challenge)
      - task:
          type: text-generation
        dataset:
          name: arc_easy
          type: arc_easy
        metrics:
          - type: Accuracy
            value: 21.4
            name: ARC (Easy)
      - task:
          type: text-generation
        dataset:
          name: gpqa
          type: gpqa
        metrics:
          - type: Accuracy
            value: 12.73
            name: GPQA
      - task:
          type: text-generation
        dataset:
          name: gsm8k
          type: gsm8k
        metrics:
          - type: Accuracy
            value: 7.73
            name: GSM8K
      - task:
          type: text-generation
        dataset:
          name: ifeval
          type: ifeval
        metrics:
          - type: Accuracy
            value: 9.01
            name: IFEVAL
      - task:
          type: text-generation
        dataset:
          name: math
          type: math
        metrics:
          - type: Accuracy
            value: 1.91
            name: MATH (Level 5)
      - task:
          type: text-generation
        dataset:
          name: mmlu
          type: mmlu
        metrics:
          - type: Accuracy
            value: 22.76
            name: MMLU
      - task:
          type: text-generation
        dataset:
          name: mt_bench
          type: mt_bench
        metrics:
          - type: Accuracy
            value: 8.2
            name: MT-Bench
      - task:
          type: text-generation
        dataset:
          name: truthful_qa
          type: truthful_qa
        metrics:
          - type: Accuracy
            value: 9.06
            name: TruthfulQA

mlx-community/lille-130m-instruct-6bit

This model mlx-community/lille-130m-instruct-6bit was converted to MLX format from mlx-community/lille-130m-instruct-fp16 using mlx-lm version 0.27.1.

Use with mlx

pip install mlx-lm

from mlx_lm import load, generate

model, tokenizer = load("mlx-community/lille-130m-instruct-6bit")

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)