gemma-2-27b-it-FP8-fix-system-role

Quantized version of gemma-2-27b-it and update chat_template for support system role to handle cases:

• Conversation roles must alternate user/assistant/user/assistant/...
• System role not supported

Model Overview

• Model Architecture: Gemma 2
  • Input: Text
  • Output: Text
• Model Optimizations:
  • Weight quantization: FP8
  • Activation quantization: FP8
• Release Date: 04/12/2024
• Version: 1.0

Model Optimizations

This model was obtained by quantizing the weights and activations of gemma-2-27b-it to FP8 data type, ready for inference with vLLM >= 0.5.1. This optimization reduces the number of bits per parameter from 16 to 8, reducing the disk size and GPU memory requirements by approximately 50%.

Only the weights and activations of the linear operators within transformers blocks are quantized. Symmetric per-tensor quantization is applied, in which a single linear scaling maps the FP8 representations of the quantized weights and activations. AutoFP8 is used for quantization with a single instance of every token in random order.

Deployment

Use with vLLM

This model can be deployed efficiently using the vLLM backend, as shown in the example below.

With CLI:

vllm serve --model dangvansam/gemma-2-27b-it-FP8-fix-system-role -q fp8

curl http://localhost:8000/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{
  "model": "dangvansam/gemma-2-27b-it-FP8-fix-system-role",
  "messages": [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": "Who are you?"}
  ]
}'

With Python:

from vllm import LLM, SamplingParams
from transformers import AutoTokenizer

model_id = "dangvansam/gemma-2-27b-it-FP8-fix-system-role"

sampling_params = SamplingParams(temperature=0.6, top_p=0.9, max_tokens=256)

tokenizer = AutoTokenizer.from_pretrained(model_id)

messages = [
  {"role": "system", "content": "You are helpfull assistant."},
  {"role": "user", "content": "Who are you?"}
]

prompts = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)

llm = LLM(model=model_id)

outputs = llm.generate(prompts, sampling_params)

generated_text = outputs[0].outputs[0].text
print(generated_text)

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