aliafshar/gemma3-it-qat-tools:12b

aliafshar/ gemma3-it-qat-tools:12b

12.6K Downloads Updated 9 months ago

Gemma3 instruction-tuned and quantized with tool/functions using a faithful representation of the Google DeepMind recommended tool calling methodology. Runs remarkably well on old GPUs and CPU.

vision tools 1b 4b 12b 27b

ollama run aliafshar/gemma3-it-qat-tools:12b

curl http://localhost:11434/api/chat \
  -d '{
    "model": "aliafshar/gemma3-it-qat-tools:12b",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='aliafshar/gemma3-it-qat-tools:12b',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'aliafshar/gemma3-it-qat-tools:12b',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Details

Updated 9 months ago

9 months ago

874cf7413552 · 8.9GB ·

model

archgemma3

parameters12.2B

quantizationQ4_0

8.9GB

template

{{- if .System }}<start_of_turn>user {{ .System }}<end_of_turn> {{- end }} {{- if .Tools }}<start_of

1.2kB

params

{ "stop": [ "<end_of_turn>" ], "temperature": 1, "top_k": 64, "top_p": 0

77B

Readme

Gemma3 instruction-tuned and quantized with tool/functions using a faithful representation of the Google DeepMind recommended tool calling methodology. Runs remarkably well on old GPUs and CPU.

References:

See the original Gemma3 information here:

https://ollama.com/library/gemma3