Instructions to use hkust-aaron/EdgeRock-4B-NVFP4 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use hkust-aaron/EdgeRock-4B-NVFP4 with Transformers:

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

pipe = pipeline("text-generation", model="hkust-aaron/EdgeRock-4B-NVFP4", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("hkust-aaron/EdgeRock-4B-NVFP4", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use hkust-aaron/EdgeRock-4B-NVFP4 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "hkust-aaron/EdgeRock-4B-NVFP4"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "hkust-aaron/EdgeRock-4B-NVFP4",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/hkust-aaron/EdgeRock-4B-NVFP4

SGLang

How to use hkust-aaron/EdgeRock-4B-NVFP4 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 "hkust-aaron/EdgeRock-4B-NVFP4" \
    --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": "hkust-aaron/EdgeRock-4B-NVFP4",
		"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 "hkust-aaron/EdgeRock-4B-NVFP4" \
        --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": "hkust-aaron/EdgeRock-4B-NVFP4",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use hkust-aaron/EdgeRock-4B-NVFP4 with Docker Model Runner:
```
docker model run hf.co/hkust-aaron/EdgeRock-4B-NVFP4
```

EdgeRock-4B-NVFP4 / hf_quant_config.json

hkust-aaron

Publish packed NVFP4 EdgeRock with verified GPU reference loading and generation

fbfb7a1 verified 13 days ago

Raw History Blame Contribute Delete

4.86 kB

{
    "producer": {
        "name": "modelopt",
        "version": "0.47.0"
    },
    "quantization": {
        "quant_algo": "W4A16_NVFP4",
        "kv_cache_quant_algo": null,
        "group_size": 16,
        "exclude_modules": [
            "lm_head",
            "model.embed_tokens",
            "model.layers.0.linear_attn.conv1d",
            "model.layers.0.linear_attn.in_proj_a",
            "model.layers.0.linear_attn.in_proj_b",
            "model.layers.1.linear_attn.conv1d",
            "model.layers.1.linear_attn.in_proj_a",
            "model.layers.1.linear_attn.in_proj_b",
            "model.layers.10.linear_attn.conv1d",
            "model.layers.10.linear_attn.in_proj_a",
            "model.layers.10.linear_attn.in_proj_b",
            "model.layers.11.memory.admission_down",
            "model.layers.11.memory.admission_up",
            "model.layers.12.linear_attn.conv1d",
            "model.layers.12.linear_attn.in_proj_a",
            "model.layers.12.linear_attn.in_proj_b",
            "model.layers.13.linear_attn.conv1d",
            "model.layers.13.linear_attn.in_proj_a",
            "model.layers.13.linear_attn.in_proj_b",
            "model.layers.14.linear_attn.conv1d",
            "model.layers.14.linear_attn.in_proj_a",
            "model.layers.14.linear_attn.in_proj_b",
            "model.layers.15.memory.admission_down",
            "model.layers.15.memory.admission_up",
            "model.layers.16.linear_attn.conv1d",
            "model.layers.16.linear_attn.in_proj_a",
            "model.layers.16.linear_attn.in_proj_b",
            "model.layers.17.linear_attn.conv1d",
            "model.layers.17.linear_attn.in_proj_a",
            "model.layers.17.linear_attn.in_proj_b",
            "model.layers.18.linear_attn.conv1d",
            "model.layers.18.linear_attn.in_proj_a",
            "model.layers.18.linear_attn.in_proj_b",
            "model.layers.19.memory.admission_down",
            "model.layers.19.memory.admission_up",
            "model.layers.2.linear_attn.conv1d",
            "model.layers.2.linear_attn.in_proj_a",
            "model.layers.2.linear_attn.in_proj_b",
            "model.layers.20.linear_attn.conv1d",
            "model.layers.20.linear_attn.in_proj_a",
            "model.layers.20.linear_attn.in_proj_b",
            "model.layers.21.linear_attn.conv1d",
            "model.layers.21.linear_attn.in_proj_a",
            "model.layers.21.linear_attn.in_proj_b",
            "model.layers.22.linear_attn.conv1d",
            "model.layers.22.linear_attn.in_proj_a",
            "model.layers.22.linear_attn.in_proj_b",
            "model.layers.23.memory.admission_down",
            "model.layers.23.memory.admission_up",
            "model.layers.24.linear_attn.conv1d",
            "model.layers.24.linear_attn.in_proj_a",
            "model.layers.24.linear_attn.in_proj_b",
            "model.layers.25.linear_attn.conv1d",
            "model.layers.25.linear_attn.in_proj_a",
            "model.layers.25.linear_attn.in_proj_b",
            "model.layers.26.linear_attn.conv1d",
            "model.layers.26.linear_attn.in_proj_a",
            "model.layers.26.linear_attn.in_proj_b",
            "model.layers.27.memory.admission_down",
            "model.layers.27.memory.admission_up",
            "model.layers.28.linear_attn.conv1d",
            "model.layers.28.linear_attn.in_proj_a",
            "model.layers.28.linear_attn.in_proj_b",
            "model.layers.29.linear_attn.conv1d",
            "model.layers.29.linear_attn.in_proj_a",
            "model.layers.29.linear_attn.in_proj_b",
            "model.layers.3.memory.admission_down",
            "model.layers.3.memory.admission_up",
            "model.layers.30.linear_attn.conv1d",
            "model.layers.30.linear_attn.in_proj_a",
            "model.layers.30.linear_attn.in_proj_b",
            "model.layers.31.memory.admission_down",
            "model.layers.31.memory.admission_up",
            "model.layers.4.linear_attn.conv1d",
            "model.layers.4.linear_attn.in_proj_a",
            "model.layers.4.linear_attn.in_proj_b",
            "model.layers.5.linear_attn.conv1d",
            "model.layers.5.linear_attn.in_proj_a",
            "model.layers.5.linear_attn.in_proj_b",
            "model.layers.6.linear_attn.conv1d",
            "model.layers.6.linear_attn.in_proj_a",
            "model.layers.6.linear_attn.in_proj_b",
            "model.layers.7.memory.admission_down",
            "model.layers.7.memory.admission_up",
            "model.layers.8.linear_attn.conv1d",
            "model.layers.8.linear_attn.in_proj_a",
            "model.layers.8.linear_attn.in_proj_b",
            "model.layers.9.linear_attn.conv1d",
            "model.layers.9.linear_attn.in_proj_a",
            "model.layers.9.linear_attn.in_proj_b"
        ]
    }
}

	{
	"producer": {
	"name": "modelopt",
	"version": "0.47.0"
	},
	"quantization": {
	"quant_algo": "W4A16_NVFP4",
	"kv_cache_quant_algo": null,
	"group_size": 16,
	"exclude_modules": [
	"lm_head",
	"model.embed_tokens",
	"model.layers.0.linear_attn.conv1d",
	"model.layers.0.linear_attn.in_proj_a",
	"model.layers.0.linear_attn.in_proj_b",
	"model.layers.1.linear_attn.conv1d",
	"model.layers.1.linear_attn.in_proj_a",
	"model.layers.1.linear_attn.in_proj_b",
	"model.layers.10.linear_attn.conv1d",
	"model.layers.10.linear_attn.in_proj_a",
	"model.layers.10.linear_attn.in_proj_b",
	"model.layers.11.memory.admission_down",
	"model.layers.11.memory.admission_up",
	"model.layers.12.linear_attn.conv1d",
	"model.layers.12.linear_attn.in_proj_a",
	"model.layers.12.linear_attn.in_proj_b",
	"model.layers.13.linear_attn.conv1d",
	"model.layers.13.linear_attn.in_proj_a",
	"model.layers.13.linear_attn.in_proj_b",
	"model.layers.14.linear_attn.conv1d",
	"model.layers.14.linear_attn.in_proj_a",
	"model.layers.14.linear_attn.in_proj_b",
	"model.layers.15.memory.admission_down",
	"model.layers.15.memory.admission_up",
	"model.layers.16.linear_attn.conv1d",
	"model.layers.16.linear_attn.in_proj_a",
	"model.layers.16.linear_attn.in_proj_b",
	"model.layers.17.linear_attn.conv1d",
	"model.layers.17.linear_attn.in_proj_a",
	"model.layers.17.linear_attn.in_proj_b",
	"model.layers.18.linear_attn.conv1d",
	"model.layers.18.linear_attn.in_proj_a",
	"model.layers.18.linear_attn.in_proj_b",
	"model.layers.19.memory.admission_down",
	"model.layers.19.memory.admission_up",
	"model.layers.2.linear_attn.conv1d",
	"model.layers.2.linear_attn.in_proj_a",
	"model.layers.2.linear_attn.in_proj_b",
	"model.layers.20.linear_attn.conv1d",
	"model.layers.20.linear_attn.in_proj_a",
	"model.layers.20.linear_attn.in_proj_b",
	"model.layers.21.linear_attn.conv1d",
	"model.layers.21.linear_attn.in_proj_a",
	"model.layers.21.linear_attn.in_proj_b",
	"model.layers.22.linear_attn.conv1d",
	"model.layers.22.linear_attn.in_proj_a",
	"model.layers.22.linear_attn.in_proj_b",
	"model.layers.23.memory.admission_down",
	"model.layers.23.memory.admission_up",
	"model.layers.24.linear_attn.conv1d",
	"model.layers.24.linear_attn.in_proj_a",
	"model.layers.24.linear_attn.in_proj_b",
	"model.layers.25.linear_attn.conv1d",
	"model.layers.25.linear_attn.in_proj_a",
	"model.layers.25.linear_attn.in_proj_b",
	"model.layers.26.linear_attn.conv1d",
	"model.layers.26.linear_attn.in_proj_a",
	"model.layers.26.linear_attn.in_proj_b",
	"model.layers.27.memory.admission_down",
	"model.layers.27.memory.admission_up",
	"model.layers.28.linear_attn.conv1d",
	"model.layers.28.linear_attn.in_proj_a",
	"model.layers.28.linear_attn.in_proj_b",
	"model.layers.29.linear_attn.conv1d",
	"model.layers.29.linear_attn.in_proj_a",
	"model.layers.29.linear_attn.in_proj_b",
	"model.layers.3.memory.admission_down",
	"model.layers.3.memory.admission_up",
	"model.layers.30.linear_attn.conv1d",
	"model.layers.30.linear_attn.in_proj_a",
	"model.layers.30.linear_attn.in_proj_b",
	"model.layers.31.memory.admission_down",
	"model.layers.31.memory.admission_up",
	"model.layers.4.linear_attn.conv1d",
	"model.layers.4.linear_attn.in_proj_a",
	"model.layers.4.linear_attn.in_proj_b",
	"model.layers.5.linear_attn.conv1d",
	"model.layers.5.linear_attn.in_proj_a",
	"model.layers.5.linear_attn.in_proj_b",
	"model.layers.6.linear_attn.conv1d",
	"model.layers.6.linear_attn.in_proj_a",
	"model.layers.6.linear_attn.in_proj_b",
	"model.layers.7.memory.admission_down",
	"model.layers.7.memory.admission_up",
	"model.layers.8.linear_attn.conv1d",
	"model.layers.8.linear_attn.in_proj_a",
	"model.layers.8.linear_attn.in_proj_b",
	"model.layers.9.linear_attn.conv1d",
	"model.layers.9.linear_attn.in_proj_a",
	"model.layers.9.linear_attn.in_proj_b"
	]
	}
	}