metadata
license: gemma
language:
- en
- zh
base_model: twinkle-ai/gemma-3-4B-T1-it
library_name: mlx
tags:
- Taiwan
- R.O.C
- zhtw
- SLM
- Gemma-3
- gemma3
- mlx
datasets:
- lianghsun/tw-reasoning-instruct
- lianghsun/tw-contract-review-chat
- minyichen/tw-instruct-R1-200k
- minyichen/tw_mm_R1
- minyichen/LongPaper_multitask_zh_tw_R1
- nvidia/Nemotron-Instruction-Following-Chat-v1
metrics:
- accuracy
pipeline_tag: text-generation
model-index:
- name: gemma-3-4B-T1-it
results:
- task:
type: question-answering
name: Single Choice Question
dataset:
name: tmmlu+
type: ikala/tmmluplus
config: all
split: test
revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c
metrics:
- type: accuracy
value: 47.44
name: single choice
- task:
type: question-answering
name: Single Choice Question
dataset:
name: mmlu
type: cais/mmlu
config: all
split: test
revision: c30699e
metrics:
- type: accuracy
value: 59.13
name: single choice
- task:
type: question-answering
name: Single Choice Question
dataset:
name: tw-legal-benchmark-v1
type: lianghsun/tw-legal-benchmark-v1
config: all
split: test
revision: 66c3a5f
metrics:
- type: accuracy
value: 44.18
name: single choice
SpockH/gemma-3-4B-T1-it-mlx-4bit
This model SpockH/gemma-3-4B-T1-it-mlx-4bit was converted to MLX format from twinkle-ai/gemma-3-4B-T1-it using mlx-lm version 0.29.1.
Use with mlx
pip install mlx-lm
from mlx_lm import load, generate
model, tokenizer = load("SpockH/gemma-3-4B-T1-it-mlx-4bit")
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)