mirror of
https://github.com/hanzoai/engine.git
synced 2026-11-20 14:17:52 +00:00
1.9 KiB
1.9 KiB
Gemma 2: collection
The Gemma 2 models are a family of text-to-text decoder-only LLMs.
Quick Start
hanzo run --isq 4 -m google/gemma-2-9b-it
HTTP API
hanzo serve --isq 4 -p 1234 -m google/gemma-2-9b-it
from openai import OpenAI
client = OpenAI(api_key="foobar", base_url="http://localhost:1234/v1/")
completion = client.chat.completions.create(
model="default",
messages=[
{"role": "user", "content": "Tell me a story about the Rust type system."}
],
max_tokens=256,
)
print(completion.choices[0].message.content)
Python SDK
from hanzo import Runner, Which, ChatCompletionRequest, Architecture
runner = Runner(
which=Which.Plain(
model_id="google/gemma-2-9b-it",
arch=Architecture.Gemma2,
),
)
res = runner.send_chat_completion_request(
ChatCompletionRequest(
model="default",
messages=[
{"role": "user", "content": "Tell me a story about the Rust type system."}
],
max_tokens=256,
presence_penalty=1.0,
top_p=0.1,
temperature=0.1,
)
)
print(res.choices[0].message.content)
print(res.usage)
Rust SDK
use anyhow::Result;
use hanzo::{
IsqType, PagedAttentionMetaBuilder, TextMessageRole, TextMessages, TextModelBuilder,
};
#[tokio::main]
async fn main() -> Result<()> {
let model = TextModelBuilder::new("google/gemma-2-9b-it")
.with_isq(IsqType::Q4K)
.with_logging()
.with_paged_attn(|| PagedAttentionMetaBuilder::default().build())?
.build()
.await?;
let messages = TextMessages::new()
.add_message(TextMessageRole::User, "Hello!");
let response = model.send_chat_request(messages).await?;
println!("{}", response.choices[0].message.content.as_ref().unwrap());
Ok(())
}