Instructions to use kaushalbalagurusamy/zamba2-2.7b-spatial-guard with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use kaushalbalagurusamy/zamba2-2.7b-spatial-guard with Transformers:

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

pipe = pipeline("text-generation", model="kaushalbalagurusamy/zamba2-2.7b-spatial-guard")
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("kaushalbalagurusamy/zamba2-2.7b-spatial-guard")
model = AutoModelForCausalLM.from_pretrained("kaushalbalagurusamy/zamba2-2.7b-spatial-guard", 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 kaushalbalagurusamy/zamba2-2.7b-spatial-guard with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "kaushalbalagurusamy/zamba2-2.7b-spatial-guard"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "kaushalbalagurusamy/zamba2-2.7b-spatial-guard",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/kaushalbalagurusamy/zamba2-2.7b-spatial-guard

SGLang

How to use kaushalbalagurusamy/zamba2-2.7b-spatial-guard 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 "kaushalbalagurusamy/zamba2-2.7b-spatial-guard" \
    --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": "kaushalbalagurusamy/zamba2-2.7b-spatial-guard",
		"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 "kaushalbalagurusamy/zamba2-2.7b-spatial-guard" \
        --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": "kaushalbalagurusamy/zamba2-2.7b-spatial-guard",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use kaushalbalagurusamy/zamba2-2.7b-spatial-guard with Docker Model Runner:
```
docker model run hf.co/kaushalbalagurusamy/zamba2-2.7b-spatial-guard
```

kaushalbalagurusamy/zamba2-2.7b-spatial-guard

Zamba2-2.7B-Instruct fine-tuned with QDoRA (NF4 4-bit base, DoRA r=16) as a multi-turn guard in front of an image-editing tool. Adapter merged into bf16 weights. Run v1-n853; dataset sha256 e448f46a343d5792e4eb55e23716d8dc35e3ed0ad837bfb39297d4512b73d5d2 (853 rows); 108 steps. Output: one JSON verdict per turn (decision allow/refuse, trigger_turn, signals, reason, tool_call). Results, limits and the data-efficiency curve: evals/RESULTS.md in https://labs.gauntletai.com/kaushalbalagurusamy/w7-small-learning-model . Needs the mamba-ssm / causal-conv1d CUDA kernels; the torch fallback gives wrong output.

Downloads last month: 631