Instructions to use smoky1496/SpawnPoint-Coder-7B-AWQ with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use smoky1496/SpawnPoint-Coder-7B-AWQ with Transformers:

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

pipe = pipeline("text-generation", model="smoky1496/SpawnPoint-Coder-7B-AWQ")
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("smoky1496/SpawnPoint-Coder-7B-AWQ")
model = AutoModelForCausalLM.from_pretrained("smoky1496/SpawnPoint-Coder-7B-AWQ", 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 smoky1496/SpawnPoint-Coder-7B-AWQ with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "smoky1496/SpawnPoint-Coder-7B-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": "smoky1496/SpawnPoint-Coder-7B-AWQ",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/smoky1496/SpawnPoint-Coder-7B-AWQ

SGLang

How to use smoky1496/SpawnPoint-Coder-7B-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 "smoky1496/SpawnPoint-Coder-7B-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": "smoky1496/SpawnPoint-Coder-7B-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 "smoky1496/SpawnPoint-Coder-7B-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": "smoky1496/SpawnPoint-Coder-7B-AWQ",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use smoky1496/SpawnPoint-Coder-7B-AWQ with Docker Model Runner:
```
docker model run hf.co/smoky1496/SpawnPoint-Coder-7B-AWQ
```

SpawnPoint-Coder-7B

A 7B model that generates Godot 4 games. Fine-tuned for Spawn Point, an open-source autonomous game studio: you describe a game in plain English or Vietnamese, the model writes the game design as strict JSON, and Spawn Point compiles it into a playable Godot 4 project (scenes, GDScript components, levels, sprites) that runs on desktop or in the browser.

What it does inside the pipeline:

Design: game idea → game brief (genre, character, abilities, enemies, hazards, pickups, goal, art style, feel).
Tweak: "jump 20% higher, add a sword, make it cartoon" → a precise patch of the game.
Edit an entity: "double its health and make it shoot fireballs" → component and parameter changes.
Translate: Vietnamese game ideas → English briefs.

4-bit AWQ, runs on a single 12 GB GPU.

What's new in v4

Farming games ("a bunny grows carrots and lettuce, moles steal them, earn $300 before sunset").
Tower defense ("orcs and trolls march to the castle, build archer towers and catapults").
Sword fighting: knights, samurai and barbarians slash up close in platformers, arenas and farms.
Cartoon art: ask for "cartoon", "hand-drawn" or "not pixel art" and Spawn Point draws smooth, non-pixel sprites, tiles and scenery instead of pixel art.
Tweaks are more reliable (held-out tweaks 84% → 100%), including removing abilities.

Run it

docker run --gpus all -p 8000:8000 --ipc=host vllm/vllm-openai:v0.10.2 \
  --model smoky1496/SpawnPoint-Coder-7B-AWQ --served-model-name SpawnPoint-Coder-7B --max-model-len 16384 --gpu-memory-utilization 0.85

Then Spawn Point finds it automatically at http://localhost:8000/v1.

Evaluation

Same prompts, schemas, normalizer and server for both versions. The sealed set for the new game kinds was written before training and not looked at until the release decision. Human-written prompts are English and Vietnamese.

benchmark	v3	v4 (this release)
sealed set for the new game kinds (12 prompts), fully right	50%	67%
16 human-written prompts for the new game kinds, fully right	50%	75%
held-out briefs for the new game kinds (exact)	32%	91%
sealed set for the older game kinds (12 prompts), fully right	67%	67%
30 human-written prompts, fully right	87%	87%
12 prompts for bubble / rescue / survivor games, fully right	75%	92%
held-out brief (exact)	85%	81%
held-out tweak (exact)	84%	100%
held-out entity (exact)	100%	100%
held-out translate (exact)	97%	97%

The 4-bit AWQ files you download, measured the same way (fully right):

benchmark	v3 AWQ	v4 AWQ
sealed set, new game kinds	33%	75%
16 prompts, new game kinds	44%	88%
30 human-written prompts	90%	90%

Limitations

Fine-tuned only for the 2D game kinds Spawn Point builds: platformers (including rescue missions and sword fighting), top-down arenas and survivors, bubble shooters, farming games and tower defense. Other genres (3D, RPG systems, card games, racing…) and free-form Godot coding were not part of the training data, so expect results no better than a general 7B coder model there.
It writes Spawn Point's JSON formats; the GDScript itself comes from Spawn Point's tested component library, not from the model. Outside Spawn Point it is a general coding model with a narrow specialty.
Occasionally one detail of a request is missed (e.g. a pickup or one enemy type); Spawn Point's validator and normalizer catch most of these. On the older held-out briefs v4 scores a little below v3 (81% vs 85% exact), mostly on the unnamed default pickup.

License

Apache 2.0, see LICENSE and NOTICE.

Downloads last month: 83

Safetensors

Model size

8B params

Tensor type

I32

BF16

F16