nm-testing/Qwen2-VL-72B-Instruct-FP8-dynamic

Qwen2-VL-72B-Instruct-quantized-w4a16

Model Overview

• Model Architecture: Qwen/Qwen2-VL-72B-Instruct
  • Input: Vision-Text
  • Output: Text
• Model Optimizations:
  • Weight quantization: FP8
  • Activation quantization: FP8
• Release Date: 2/24/2025
• Version: 1.0
• Model Developers: Neural Magic

Quantized version of Qwen/Qwen2-VL-72B-Instruct.

Model Optimizations

This model was obtained by quantizing the weights of Qwen/Qwen2-VL-72B-Instruct to FP8 data type, ready for inference with vLLM >= 0.5.2.

Deployment

Use with vLLM

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

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

# prepare model
llm = LLM(
    model="neuralmagic/Qwen2-VL-72B-Instruct-quantized.w4a16",
    trust_remote_code=True,
    max_model_len=4096,
    max_num_seqs=2,
)

# prepare inputs
question = "What is the content of this image?"
inputs = {
    "prompt": f"<|user|>\n<|image_1|>\n{question}<|end|>\n<|assistant|>\n",
    "multi_modal_data": {
        "image": ImageAsset("cherry_blossom").pil_image.convert("RGB")
    },
}

# generate response
print("========== SAMPLE GENERATION ==============")
outputs = llm.generate(inputs, SamplingParams(temperature=0.2, max_tokens=64))
print(f"PROMPT  : {outputs[0].prompt}")
print(f"RESPONSE: {outputs[0].outputs[0].text}")
print("==========================================")

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

Creation

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

Model Creation Code

from transformers import AutoProcessor, Qwen2VLForConditionalGeneration

from llmcompressor.modifiers.quantization import QuantizationModifier
from llmcompressor.transformers import oneshot, wrap_hf_model_class

MODEL_ID = "Qwen/Qwen2-VL-72B-Instruct"

# Load model.
model_class = wrap_hf_model_class(Qwen2VLForConditionalGeneration)
model = model_class.from_pretrained(MODEL_ID, device_map="auto", torch_dtype="auto")
processor = AutoProcessor.from_pretrained(MODEL_ID)

# Configure the quantization algorithm and scheme.
# In this case, we:
#   * quantize the weights to fp8 with per channel via ptq
#   * quantize the activations to fp8 with dynamic per token
recipe = QuantizationModifier(
    targets="Linear",
    scheme="FP8_DYNAMIC",
    ignore=["re:.*lm_head", "re:visual.*"],
)

# Apply quantization and save to disk in compressed-tensors format.
SAVE_DIR = MODEL_ID.split("/")[1] + "-FP8-dynamic"
oneshot(model=model, recipe=recipe, output_dir=SAVE_DIR)
processor.save_pretrained(SAVE_DIR)

# Confirm generations of the quantized model look sane.
print("========== SAMPLE GENERATION ==============")
input_ids = processor(text="Hello my name is", return_tensors="pt").input_ids.to("cuda")
output = model.generate(input_ids, max_new_tokens=20)
print(processor.decode(output[0]))
print("==========================================")

Evaluation

The model was evaluated using mistral-evals for vision-related tasks and using lm_evaluation_harness for select text-based benchmarks. The evaluations were conducted using the following commands:

Evaluation Commands

Vision Tasks

• vqav2
• docvqa
• mathvista
• mmmu
• chartqa

vllm serve neuralmagic/pixtral-12b-quantized.w8a8 --tensor_parallel_size 1 --max_model_len 25000 --trust_remote_code --max_num_seqs 8 --gpu_memory_utilization 0.9 --dtype float16 --limit_mm_per_prompt image=7

python -m eval.run eval_vllm \
        --model_name neuralmagic/pixtral-12b-quantized.w8a8 \
        --url http://0.0.0.0:8000 \
        --output_dir ~/tmp \
        --eval_name <vision_task_name>

Text-based Tasks

MMLU

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 \
  --tasks mmlu \
  --num_fewshot 5 \
  --batch_size auto \
  --output_path output_dir

MGSM

lm_eval \
  --model vllm \
  --model_args pretrained="<model_name>",dtype=auto,max_model_len=4096,max_gen_toks=2048,max_num_seqs=128,tensor_parallel_size=<n>,gpu_memory_utilization=0.9 \
  --tasks mgsm_cot_native \
  --num_fewshot 0 \
  --batch_size auto \
  --output_path output_dir

Accuracy

Category	Metric	Qwen/Qwen2-VL-72B-Instruct	neuralmagic/Qwen2-VL-72B-Instruct-FP8-Dynamic	Recovery (%)
Vision	MMMU (val, CoT) explicit_prompt_relaxed_correctness	62.11	60.67	97.68%
	VQAv2 (val) vqa_match	82.51	82.44	99.91%
	DocVQA (val) anls	95.01	95.10	100.09%
	ChartQA (test, CoT) anywhere_in_answer_relaxed_correctness	83.40	83.68	100.34%
	Mathvista (testmini, CoT) explicit_prompt_relaxed_correctness	66.57	67.07	100.75%
	Average Score	77.12	77.39	100.35%
Text	MGSM (CoT)	68.60	67.78	98.80%
	MMLU (5-shot)	82.70	82.60	99.88%

Inference Performance

This model achieves up to xxx speedup in single-stream deployment and up to xxx speedup in multi-stream asynchronous deployment, depending on hardware and use-case scenario. The following performance benchmarks were conducted with vLLM version 0.7.2, and GuideLLM.

Benchmarking Command

``` guidellm --model neuralmagic/Qwen2-VL-72B-Instruct-FP8-Dynamic --target "http://localhost:8000/v1" --data-type emulated --data prompt_tokens=,generated_tokens=,images=,width=,height= --max seconds 120 --backend aiohttp_server ```

Single-stream performance (measured with vLLM version 0.7.2)

			Document Visual Question Answering 1680W x 2240H 64/128		Visual Reasoning 640W x 480H 128/128		Image Captioning 480W x 360H 0/128
Hardware	Number of GPUs	Model	Average Cost Reduction	Latency (s)	QPD	Latency (s)th>	QPD	Latency (s)	QPD
A100	4	Qwen/Qwen2-VL-72B-Instruct		6.5	77	4.6	110	4.4	113
	2	neuralmagic/Qwen2-VL-72B-Instruct-quantized.w8a8	1.85	7.2	139	4.9	206	4.8	211
	1	neuralmagic/Qwen2-VL-72B-Instruct-quantized.w4a16	3.32	10.0	202	5.0	398	4.8	419
H100	4	Qwen/Qwen2-VL-72B-Instruct		4.4	66	3.0	97	2.9	99
	2	neuralmagic/Qwen2-VL-72B-Instruct-FP8-Dynamic	1.79	4.7	119	3.3	173	3.2	177
	1	neuralmagic/Qwen2-VL-72B-Instruct-quantized.w4a16	2.60	6.4	172	4.3	253	4.2	259

**Use case profiles: Image Size (WxH) / prompt tokens / generation tokens

**QPD: Queries per dollar, based on on-demand cost at Lambda Labs (observed on 2/18/2025).

Multi-stream asynchronous performance (measured with vLLM version 0.7.2)

			Document Visual Question Answering 1680W x 2240H 64/128		Visual Reasoning 640W x 480H 128/128		Image Captioning 480W x 360H 0/128
Hardware	Model	Average Cost Reduction	Maximum throughput (QPS)	QPD	Maximum throughput (QPS)	QPD	Maximum throughput (QPS)	QPD
A100x4	Qwen/Qwen2-VL-72B-Instruct		0.3	169	1.1	538	1.2	595
	neuralmagic/Qwen2-VL-72B-Instruct-quantized.w8a8	1.84	1.2	586	4.0	2042	4.6	2270
	neuralmagic/Qwen2-VL-72B-Instruct-quantized.w4a16	2.73	2.4	1256	12.8	6364	16.0	8076
H100x4	Qwen/Qwen2-VL-72B-Instruct		0.5	137	1.2	356	1.3	377
	neuralmagic/Qwen2-VL-72B-Instruct-FP8-Dynamic	1.70	1.6	457	4.4	1207	4.8	1296
	neuralmagic/Qwen2-VL-72B-Instruct-quantized.w4a16	2.35	5.2	1400	13.2	3640	14.4	3976

**Use case profiles: Image Size (WxH) / prompt tokens / generation tokens

**QPS: Queries per second.

**QPD: Queries per dollar, based on on-demand cost at Lambda Labs (observed on 2/18/2025).