Gemma 3 27b It FP8 Dynamic

🚀 gemma-3-27b-it-FP8-Dynamic

This is a quantized version of google/gemma-3-27b-it, which offers optimized performance with specific quantization techniques and is suitable for efficient inference with vLLM.

🚀 Quick Start

The model can be quickly deployed using the vLLM backend. Refer to the Deployment section for detailed code examples.

✨ Features

• Multimodal Input: Accepts vision - text as input and generates text output.
• Quantization Optimization: Both weight and activation are quantized to FP8 data type, enabling efficient inference.
• vLLM Compatibility: Supports deployment with vLLM >= 0.5.2 and OpenAI - compatible serving.

📦 Installation

No specific installation steps are provided in the original document.

💻 Usage Examples

Basic Usage

This model can be deployed efficiently using the vLLM backend.

from vllm import LLM, SamplingParams
from vllm.assets.image import ImageAsset
from transformers import AutoProcessor

# Define model name once
model_name = "RedHatAI/gemma-3-27b-it-FP8-dynamic"

# Load image and processor
image = ImageAsset("cherry_blossom").pil_image.convert("RGB")
processor = AutoProcessor.from_pretrained(model_name, trust_remote_code=True)

# Build multimodal prompt
chat = [
    {"role": "user", "content": [{"type": "image"}, {"type": "text", "text": "What is the content of this image?"}]},
    {"role": "assistant", "content": []}
]
prompt = processor.apply_chat_template(chat, add_generation_prompt=True)

# Initialize model
llm = LLM(model=model_name, trust_remote_code=True)

# Run inference
inputs = {"prompt": prompt, "multi_modal_data": {"image": [image]}}
outputs = llm.generate(inputs, SamplingParams(temperature=0.2, max_tokens=64))

# Display result
print("RESPONSE:", outputs[0].outputs[0].text)

vLLM also supports OpenAI - compatible serving. See the documentation for more details.

📚 Documentation

Model Overview

This model was obtained by quantizing the weights of google/gemma-3-27b-it to FP8 data type, ready for inference with vLLM >= 0.5.2.

Deployment

Use with vLLM

The model can be deployed efficiently using the vLLM backend. The above code example demonstrates the basic process.

Creation

This model was created with llm - compressor by running the following code snippet as part of a multimodal announcement blog.

import requests
import torch
from PIL import Image
from transformers import AutoProcessor, Gemma3ForConditionalGeneration
from llmcompressor.transformers import oneshot
from llmcompressor.modifiers.quantization import QuantizationModifier

# Load model.
model_id = google/gemma-3-27b-it
model = Gemma3ForConditionalGeneration.from_pretrained(
    model_id, device_map="auto", torch_dtype="auto"
)
processor = AutoProcessor.from_pretrained(model_id, trust_remote_code=True)

# Recipe
recipe = [
    QuantizationModifier(
        targets="Linear",
        scheme="FP8_DYNAMIC",
        sequential_targets=["Gemma3DecoderLayer"],
        ignore=["re:.*lm_head", "re:vision_tower.*", "re:multi_modal_projector.*"],
    ),
]

SAVE_DIR=f"{model_id.split('/')[1]}-FP8-Dynamic"

# Perform oneshot
oneshot(
    model=model,
    recipe=recipe,
    trust_remote_code_model=True,
    output_dir=SAVE_DIR
)

Evaluation

The model was evaluated using lm_evaluation_harness for OpenLLM v1 text benchmark. The evaluations were conducted using the following commands:

lm_eval \
  --model vllm \
  --model_args pretrained="<model_name>",dtype=auto,add_bos_token=True,max_model_len=4096,tensor_parallel_size=<n>,gpu_memory_utilization=0.8,enable_chunked_prefill=True,trust_remote_code=True,enforce_eager=True \
  --tasks openllm \
  --batch_size auto

Accuracy

📄 License

This project is licensed under the Apache 2.0 License.