Instructions to use caiovicentino1/Qwen3.5-9B-EOQ-v3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use caiovicentino1/Qwen3.5-9B-EOQ-v3 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="caiovicentino1/Qwen3.5-9B-EOQ-v3")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("caiovicentino1/Qwen3.5-9B-EOQ-v3")
model = AutoModelForMultimodalLM.from_pretrained("caiovicentino1/Qwen3.5-9B-EOQ-v3")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use caiovicentino1/Qwen3.5-9B-EOQ-v3 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "caiovicentino1/Qwen3.5-9B-EOQ-v3"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "caiovicentino1/Qwen3.5-9B-EOQ-v3",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/caiovicentino1/Qwen3.5-9B-EOQ-v3

SGLang

How to use caiovicentino1/Qwen3.5-9B-EOQ-v3 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "caiovicentino1/Qwen3.5-9B-EOQ-v3" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "caiovicentino1/Qwen3.5-9B-EOQ-v3",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "caiovicentino1/Qwen3.5-9B-EOQ-v3" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "caiovicentino1/Qwen3.5-9B-EOQ-v3",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use caiovicentino1/Qwen3.5-9B-EOQ-v3 with Docker Model Runner:
```
docker model run hf.co/caiovicentino1/Qwen3.5-9B-EOQ-v3
```

🏆 EOQ v3 -- Qwen3.5-9B (PolarQuant + AWQ)

Near-lossless quantization: PPL 6.43 -- only +0.06 from FP16 6.37. The best quality result in the PolarQuant family.

EOQ v3 combines PolarQuant (Hadamard + Lloyd-Max) with AWQ (Activation-Aware Weight Quantization) to achieve 93% reduction in quantization error vs standard absmax. This is practically indistinguishable from the full-precision model.

🎯 Key Results

Metric	Value
Method	PolarQuant Q5 + AWQ
Perplexity (WikiText-2)	6.43
FP16 Baseline	6.37
Delta from FP16	+0.06 (near-lossless!)
Download Size	~5 GB (3.6x compression)
Load Time	9s (5x faster than FP16's 53s)
Throughput	45.8 tok/s (identical to FP16)
GPU Dequant	3.5s (one-time)

📊 Quality Evolution

Version	Technique	PPL	Delta	Improvement
v1	Absmax uniform Q5	7.31	+0.94	Baseline
v2	AWQ + mixed-bit	7.05	+0.68	28% better
v3	PolarQuant + AWQ	6.43	+0.06	94% better

From v1 to v3: 93% reduction in quality loss (0.94 -> 0.06 PPL delta). PolarQuant + AWQ is the key combination.

Cross-Model Results

Model	FP16 PPL	EOQ v3 PPL	Delta
Qwen3.5-9B	6.37	6.43	+0.06
Qwen3.5-35B-A3B (MoE)	5.19	5.36	+0.17

🔬 How It Works

EOQ v3 combines two complementary techniques:

1. AWQ (Activation-Aware Scaling)

Protects important weight channels by pre-scaling them before quantization. Channels that carry more activation energy get higher precision.

2. PolarQuant (Hadamard + Lloyd-Max)

Transforms weight blocks to Gaussian via Hadamard rotation, then applies MSE-optimal Lloyd-Max quantization.

                              AWQ Pre-Scaling
                                    |
                                    v
Original Weights --> Scale Important Channels --> Normalize --> Hadamard Rotate
                                                                      |
                                                                      v
                                                            Lloyd-Max Quantize
                                                                      |
                                                                      v
                                                               Store Codes +
                                                              AWQ Scales +
                                                             Block Norms +
                                                            Centroid Table

Why They Combine Well

AWQ operates on channels (column-level scaling)
PolarQuant operates on blocks (128-element sub-vectors)
They address orthogonal sources of error: AWQ handles channel sensitivity, PolarQuant handles within-block distribution

🚀 Quick Start

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained(
    "caiovicentino1/Qwen3.5-9B-EOQ-v3",
    dtype="bfloat16", device_map="auto", trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained("caiovicentino1/Qwen3.5-9B-EOQ-v3")

output = model.generate(
    **tokenizer("Write a detailed explanation of neural network quantization:", return_tensors="pt").to("cuda"),
    max_new_tokens=300
)
print(tokenizer.decode(output[0], skip_special_tokens=True))

With torchao INT4 (for maximum speed)

from torchao.quantization import quantize_, Int4WeightOnlyConfig

# After loading EOQ v3 model (already dequanted to BF16):
quantize_(model, Int4WeightOnlyConfig(group_size=128))
# Now runs at 43+ tok/s with 6.5 GB VRAM

🔧 Technical Details

Component	Details
Quantization	PolarQuant Q5 + AWQ (5-bit, block_size=128)
AWQ	Activation-aware per-channel scaling (FP16 scales stored)
Rotation	128x128 Walsh-Hadamard (self-inverse, deterministic)
Centroids	Pre-computed MSE-optimal for N(0,1), stored in metadata (no scipy needed)
Storage	Bit-packed uint8 codes + fp16 norms + fp16 AWQ scales + fp32 centroids
GPU Dequant	unpack -> centroid lookup -> inverse Hadamard -> scale by norm -> undo AWQ
Dequant Time	3.5s (100x faster than CPU numpy)
Compression	3.6x (17.9 GB -> ~5 GB)

Storage Format

{layer_name}.packed     -- bit-packed uint8 quantization codes
{layer_name}.norms      -- fp16 per-block normalization factors
{layer_name}.awq_scales -- fp16 per-channel AWQ importance scales
metadata:
  centroids             -- fp32 Lloyd-Max optimal centroid table (shared)
  bits_per_tensor       -- quantization bits (5 for Q5)

📊 Ablation

: Why Both Components Matter

Configuration	PPL	Delta
Absmax Q5 (baseline)	7.31	+0.94
AWQ only	7.05	+0.68
PolarQuant only	6.56	+0.19
PolarQuant + AWQ	6.43	+0.06

AWQ alone reduces error by 28%. PolarQuant alone reduces error by 80%. Together they reduce error by 94% -- the effects are complementary, not redundant.

🔗 Links

\U0001f4c4 Paper (arXiv) -- PolarQuant: Optimal Gaussian Weight Quantization
💻 Code (GitHub) -- Full research codebase
\U0001f50c vLLM Plugin -- Production inference integration
🧊 PolarQuant Q5 -- Without AWQ (simpler, slightly less quality)
📊 35B MoE Version -- PPL 5.36, 4.44x compression

📖 Citation

@article{vicentino2026polarquant,
  title={PolarQuant: Optimal Gaussian Weight Quantization via Hadamard Rotation for LLM Compression},
  author={Vicentino, Caio},
  journal={arXiv preprint arXiv:2603.7424577},
  year={2026}
}

🙏 Acknowledgements

Built with PyTorch, torchao, AWQ methodology from MIT HAN Lab, and the Qwen team's open-weight models.

Downloads last month: 62

Safetensors

Model size

5B params

Tensor type

F16

Model tree for caiovicentino1/Qwen3.5-9B-EOQ-v3

Base model

Qwen/Qwen3.5-9B-Base

Finetuned

Qwen/Qwen3.5-9B

Finetuned

(382)

this model

Collections including caiovicentino1/Qwen3.5-9B-EOQ-v3