Instructions to use musatur/Alpona-1.0-9B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use musatur/Alpona-1.0-9B with Transformers:

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

pipe = pipeline("text-generation", model="musatur/Alpona-1.0-9B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("musatur/Alpona-1.0-9B")
model = AutoModelForCausalLM.from_pretrained("musatur/Alpona-1.0-9B", device_map="auto")
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=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use musatur/Alpona-1.0-9B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "musatur/Alpona-1.0-9B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "musatur/Alpona-1.0-9B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/musatur/Alpona-1.0-9B

SGLang

How to use musatur/Alpona-1.0-9B 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 "musatur/Alpona-1.0-9B" \
    --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": "musatur/Alpona-1.0-9B",
		"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 "musatur/Alpona-1.0-9B" \
        --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": "musatur/Alpona-1.0-9B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use musatur/Alpona-1.0-9B with Docker Model Runner:
```
docker model run hf.co/musatur/Alpona-1.0-9B
```

Alpona-1.0-9B

A 9B RTL generation model: Qwen3.5-9B post-trained for SystemVerilog design generation.

This repository holds the complete merged weights — no adapter, no external base to resolve, no checkpoint directories. Download it and it runs.

Results

benchmark	pass@1	pass@5
VerilogEval v2 (156 spec-to-RTL tasks)	74.3%	79.2%
RTLLM v2 (50 designs)	66.7%	73.4%

Scored with the official testbenches for each benchmark.

Thinking mode is required

The prompt prefix ends with an open <think>. The chat template in this repo does that for you via add_generation_prompt=True. Scoring an Alpona model with the reasoning block closed costs it a large fraction of its score — do not disable thinking.

Requirements

This is a hybrid architecture (model_type: qwen3_5_text). It will not load on older transformers:

pip install -U "transformers>=5.14.1" torch accelerate

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "musatur/Alpona-1.0-9B"
tok = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, dtype="bfloat16", device_map="auto")

messages = [
    {"role": "system", "content": "You are a synthesizable SystemVerilog generation agent. "
                                  "Return only one complete module with the requested interface."},
    {"role": "user", "content": "Design a 4-bit synchronous up counter named TopModule with "
                                "inputs clk, reset and output [3:0] q."},
]
prompt = tok.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
out = model.generate(**tok(prompt, return_tensors="pt").to(model.device), max_new_tokens=8192)
print(tok.decode(out[0], skip_special_tokens=True))

Licence

other / unresolved-upstream, inherited from the training corpus.

Downloads last month: 8

Safetensors

Model size

9B params

Tensor type

BF16

Model tree for musatur/Alpona-1.0-9B

Base model

Qwen/Qwen3.5-9B-Base

Finetuned

Qwen/Qwen3.5-9B

Finetuned

(1207)

this model