Gemma 4开放的力量

# Linux
curl -fsSL https://ollama.com/install.sh | sh

# macOS
brew install ollama

# Windows：从 ollama.com 下载安装包

# 默认模型（E4B，9.6GB）
ollama run gemma4

# 指定模型
ollama run gemma4:e2b    # 边缘模型，7.2GB
ollama run gemma4:26b    # MoE 模型，18GB
ollama run gemma4:31b    # 旗舰模型，20GB

# 图片理解
ollama run gemma4 "描述这张图片 /path/to/photo.png"

from openai import OpenAI

client = OpenAI(base_url="http://localhost:11434/v1", api_key="ollama")

response = client.chat.completions.create(
    model="gemma4",
    messages=[{"role": "user", "content": "你好"}]
)
print(response.choices[0].message.content)

pip install -U transformers torch accelerate
pip install bitsandbytes  # 可选，用于量化

from transformers import pipeline

pipe = pipeline("any-to-any", model="google/gemma-4-e4b-it", device_map="auto")

messages = [{
    "role": "user",
    "content": [{"type": "text", "text": "解释量子计算"}],
}]

output = pipe(messages, max_new_tokens=200, return_full_text=False)
print(output[0]["generated_text"])

messages = [{
    "role": "user",
    "content": [
        {"type": "image", "image": "https://example.com/photo.jpg"},
        {"type": "text", "text": "描述这张图片"},
    ],
}]

output = pipe(messages, max_new_tokens=200, return_full_text=False)

from transformers import AutoModelForImageTextToText, BitsAndBytesConfig

config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_compute_dtype=torch.bfloat16,
)

model = AutoModelForImageTextToText.from_pretrained(
    "google/gemma-4-31b-it",
    quantization_config=config,
    device_map="auto",
)

inputs = processor.apply_chat_template(
    messages, tokenize=True, return_dict=True,
    return_tensors="pt", add_generation_prompt=True,
    enable_thinking=True,  # 开启思维链
).to(model.device)

pip install -U vllm
pip install transformers==5.5.0

# E4B 单卡
vllm serve google/gemma-4-E4B-it \
    --max-model-len 131072

# 26B MoE 单卡（24GB+ VRAM）
vllm serve google/gemma-4-26B-A4B-it \
    --max-model-len 32768 \
    --gpu-memory-utilization 0.90

# 31B 旗舰双卡并行
vllm serve google/gemma-4-31B-it \
    --tensor-parallel-size 2 \
    --max-model-len 32768

from openai import OpenAI

client = OpenAI(base_url="http://localhost:8000/v1", api_key="EMPTY")

response = client.chat.completions.create(
    model="google/gemma-4-26B-A4B-it",
    messages=[{"role": "user", "content": "你好"}],
    max_tokens=512,
    temperature=0.7
)

vllm serve nvidia/Gemma-4-31B-IT-NVFP4 \
    --quantization modelopt \
    --tensor-parallel-size 8

git clone https://github.com/ggml-org/llama.cpp
cmake llama.cpp -B llama.cpp/build \
    -DBUILD_SHARED_LIBS=OFF -DGGML_CUDA=ON
cmake --build llama.cpp/build --config Release -j

# 小模型用 Q8_0 量化
./llama.cpp/build/bin/llama-cli \
    -hf unsloth/gemma-4-E4B-it-GGUF:Q8_0 \
    --temp 1.0 --top-p 0.95 --top-k 64

# 大模型用 Dynamic 4-bit 量化
./llama.cpp/build/bin/llama-cli \
    -hf unsloth/gemma-4-26B-A4B-it-GGUF:UD-Q4_K_XL \
    --temp 1.0 --top-p 0.95 --top-k 64

# 31B 旗舰
./llama.cpp/build/bin/llama-cli \
    -hf unsloth/gemma-4-31B-it-GGUF:UD-Q4_K_XL \
    --temp 1.0 --top-p 0.95 --top-k 64

./llama.cpp/build/bin/llama-server \
    -hf unsloth/gemma-4-26B-A4B-it-GGUF:UD-Q4_K_XL \
    --temp 1.0 --top-p 0.95 --top-k 64 \
    --chat-template-kwargs '{"enable_thinking":true}'

from google import genai

client = genai.Client(api_key="YOUR_API_KEY")

response = client.models.generate_content(
    model="gemma-4-31b-it",
    contents="用一句话解释相对论",
)
print(response.text)

// build.gradle
implementation("com.google.ai.edge.litert:litert-lm:latest")

// 初始化并推理
val model = LlmInference.create(context, "gemma-4-e2b-it")
val result = model.generateResponse("你好")