Instructions to use ree2raz/CyberSecQwen-4B-AWQ with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ree2raz/CyberSecQwen-4B-AWQ with Transformers:

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

pipe = pipeline("text-generation", model="ree2raz/CyberSecQwen-4B-AWQ")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("ree2raz/CyberSecQwen-4B-AWQ")
model = AutoModelForCausalLM.from_pretrained("ree2raz/CyberSecQwen-4B-AWQ")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.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(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use ree2raz/CyberSecQwen-4B-AWQ with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ree2raz/CyberSecQwen-4B-AWQ"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ree2raz/CyberSecQwen-4B-AWQ",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/ree2raz/CyberSecQwen-4B-AWQ

SGLang

How to use ree2raz/CyberSecQwen-4B-AWQ 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 "ree2raz/CyberSecQwen-4B-AWQ" \
    --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": "ree2raz/CyberSecQwen-4B-AWQ",
		"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 "ree2raz/CyberSecQwen-4B-AWQ" \
        --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": "ree2raz/CyberSecQwen-4B-AWQ",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use ree2raz/CyberSecQwen-4B-AWQ with Docker Model Runner:
```
docker model run hf.co/ree2raz/CyberSecQwen-4B-AWQ
```

ree2raz commited on 9 days ago

Commit

3329f36

verified ·

1 Parent(s): ec12b47

Update eval with GGUF comparison; RCM 0.5814, MCQ 0.5921

Browse files

Files changed (1) hide show

README.md +28 -18

README.md CHANGED Viewed

@@ -18,10 +18,6 @@ pipeline_tag: text-generation
 4-bit AWQ quantized version of [CyberSecQwen-4B](https://huggingface.co/lablab-ai-amd-developer-hackathon/CyberSecQwen-4B).
-## Evaluation Infrastructure
-[GitHub repository](https://github.com/ree2raz/cyberSecQwen_4b_4bit) — Modal scripts for AWQ quantization + vLLM CTI-Bench evaluation.
 ## Quantization
 | Parameter | Value |
@@ -35,19 +31,20 @@ pipeline_tag: text-generation
 ## CTI-Bench Evaluation
-Evaluated under the [Foundation-Sec-8B protocol](https://arxiv.org/abs/2504.21039) (arXiv:2504.21039):
 - Temperature 0.3, max_tokens 512, concurrency 32
 - 5 independent trials, zero-shot (no system prompt)
 - vLLM v0.20.1 with awq_marlin kernel on Modal L4 GPU
-| Task | AWQ 4-bit | FP16 Reference | Delta |
 |---|---|---|---|
-| CTI-MCQ (2,500 items) | 0.5921 +/- 0.0083 | 0.5868 +/- 0.0029 | +0.0053 |
-| CTI-RCM (1,000 items) | 0.5814 +/- 0.0025 | 0.6664 +/- 0.0023 | -0.0850 |
 **Key findings:**
-- **CTI-MCQ**: AWQ 4-bit matches or slightly exceeds FP16 performance (+0.5 points). No measurable accuracy loss.
-- **CTI-RCM**: AWQ 4-bit degrades by 8.5 percentage points vs FP16. Parseable rate > 99.8% so answer extraction is working correctly. The model retains correct CWE identification in reasoning but sometimes diverges on final answers. This gap can likely be reduced with more calibration data.
 ## Trial results
@@ -62,13 +59,22 @@ Evaluated under the [Foundation-Sec-8B protocol](https://arxiv.org/abs/2504.2103
 ### CTI-RCM
 | Trial | Seed | Accuracy |
-|-------|------|----------|
 | 1 | 42 | 0.5790 |
 | 2 | 43 | 0.5830 |
 | 3 | 44 | 0.5790 |
 | 4 | 45 | 0.5840 |
 | 5 | 46 | 0.5820 |
 ## Usage with vLLM
 ```bash
@@ -85,11 +91,15 @@ vllm serve ree2raz/CyberSecQwen-4B-AWQ --quantization awq_marlin --dtype float16
 ## Citation
 ```bibtex
-@misc{cybersecqwen2026,
-  title  = {CyberSecQwen-4B: A Compact CTI Specialist Fine-Tuned from Qwen3-4B-Instruct-2507 on AMD MI300X},
-  author = {Mulia, Samuel},
-  year   = {2026},
-  publisher = {Hugging Face},
-  url    = {https://huggingface.co/athena129/CyberSecQwen-4B}
-}
 ```

 4-bit AWQ quantized version of [CyberSecQwen-4B](https://huggingface.co/lablab-ai-amd-developer-hackathon/CyberSecQwen-4B).
 ## Quantization
 | Parameter | Value |
 ## CTI-Bench Evaluation
+Evaluated under the [Foundation-Sec-8B protocol](https://arxiv.org/abs/2504.21039):
 - Temperature 0.3, max_tokens 512, concurrency 32
 - 5 independent trials, zero-shot (no system prompt)
 - vLLM v0.20.1 with awq_marlin kernel on Modal L4 GPU
+| Task | AWQ 4-bit | GGUF Q4_K_M | FP16 Reference |
 |---|---|---|---|
+| CTI-MCQ (2,500 items) | **0.5921** ± 0.0083 | 0.5368 ± 0.0048 | 0.5868 ± 0.0029 |
+| CTI-RCM (1,000 items) | 0.5814 ± 0.0025 | **0.6254 ± 0.0063** | 0.6664 ± 0.0023 |
 **Key findings:**
+- **CTI-MCQ**: AWQ 4-bit matches or slightly exceeds FP16 performance (+0.5 points). Better than GGUF Q4_K_M.
+- **CTI-RCM**: AWQ 4-bit degrades by 8.5 percentage points vs FP16. GGUF Q4_K_M does better on this task (-4.1 pts).
+- AWQ is best for MCQ (general language), GGUF is best for RCM (task-specific classification).
 ## Trial results
 ### CTI-RCM
 | Trial | Seed | Accuracy |
+|---|---|
 | 1 | 42 | 0.5790 |
 | 2 | 43 | 0.5830 |
 | 3 | 44 | 0.5790 |
 | 4 | 45 | 0.5840 |
 | 5 | 46 | 0.5820 |
+## Quantization variants
+| Variant | CTI-MCQ | CTI-RCM | Size | Engine |
+|---|---|---|---|---|
+| [AWQ 4-bit](https://huggingface.co/ree2raz/CyberSecQwen-4B-AWQ) | 0.5921 | 0.5814 | 2.7 GB | vLLM |
+| [GGUF Q4_K_M](https://huggingface.co/ree2raz/CyberSecQwen-4B-GGUF) | 0.5368 | 0.6254 | 2.5 GB | llama.cpp |
+Choose AWQ for MCQ/general chat, GGUF for vulnerability classification.
 ## Usage with vLLM
 ```bash
 ## Citation
 ```bibtex
+@misc{{cybersecqwen2026,
+  title  = {{CyberSecQwen-4B: A Compact CTI Specialist Fine-Tuned from Qwen3-4B-Instruct-2507 on AMD MI300X}},
+  author = {{Mulia, Samuel}},
+  year   = {{2026}},
+  publisher = {{Hugging Face}},
+  url    = {{https://huggingface.co/athena129/CyberSecQwen-4B}}
+}}
 ```
+## Evaluation Infrastructure
+[GitHub repository](https://github.com/ree2raz/cyberSecQwen_4b_4bit) — Modal scripts for quantization + evaluation.