Instructions to use Project-Carbon/dcc-help-gemma-4-31b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries
PEFT
How to use Project-Carbon/dcc-help-gemma-4-31b with PEFT:
```
Task type is invalid.
```

How to use Project-Carbon/dcc-help-gemma-4-31b with Transformers:

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

pipe = pipeline("text-generation", model="Project-Carbon/dcc-help-gemma-4-31b")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoProcessor, AutoModelForImageTextToText

processor = AutoProcessor.from_pretrained("Project-Carbon/dcc-help-gemma-4-31b")
model = AutoModelForImageTextToText.from_pretrained("Project-Carbon/dcc-help-gemma-4-31b")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.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(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Project-Carbon/dcc-help-gemma-4-31b with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Project-Carbon/dcc-help-gemma-4-31b"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Project-Carbon/dcc-help-gemma-4-31b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Project-Carbon/dcc-help-gemma-4-31b

SGLang

How to use Project-Carbon/dcc-help-gemma-4-31b 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 "Project-Carbon/dcc-help-gemma-4-31b" \
    --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": "Project-Carbon/dcc-help-gemma-4-31b",
		"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 "Project-Carbon/dcc-help-gemma-4-31b" \
        --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": "Project-Carbon/dcc-help-gemma-4-31b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Unsloth Studio

How to use Project-Carbon/dcc-help-gemma-4-31b with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for Project-Carbon/dcc-help-gemma-4-31b to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for Project-Carbon/dcc-help-gemma-4-31b to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for Project-Carbon/dcc-help-gemma-4-31b to start chatting

Load model with FastModel

pip install unsloth
from unsloth import FastModel
model, tokenizer = FastModel.from_pretrained(
    model_name="Project-Carbon/dcc-help-gemma-4-31b",
    max_seq_length=2048,
)

Docker Model Runner
How to use Project-Carbon/dcc-help-gemma-4-31b with Docker Model Runner:
```
docker model run hf.co/Project-Carbon/dcc-help-gemma-4-31b
```

dcc-help-gemma-4-31b

Developed by: Juno AI
License: apache-2.0
Finetuned from model: unsloth/gemma-4-31b-it
Method: LoRA fine-tune via Unsloth + Huggingface's TRL library

This is a dcc-help model: a LoRA fine-tune of unsloth/gemma-4-31b-it, trained on the dcc-help data together with the EdStem Forum data.

What is dcc-help?

dcc-help is a terminal-based coding help assistant for C programming. When a student hits an error while compiling or running their C program (via dcc, UNSW CSE's Debugging C Compiler), they can run dcc-help to get an AI-generated explanation of what went wrong.

The assistant is pedagogical: it explains the error and guides the student toward understanding it rather than handing over a ready-made fix. The goal is to help students learn to debug their own code.

Model details

Base model: unsloth/gemma-4-31b-it
Fine-tuning: LoRA (PEFT) via Unsloth + TRL (SFT)
Training data: specific course/forum corpus
Task: Explaining C compiler/runtime errors to students

Intended use

Powers the dcc-help assistant for UNSW CSE C-programming courses. Given a student's code and the associated compiler/runtime error, it produces an explanatory, guidance-oriented response that helps the student diagnose the problem themselves.

Limitations

"No direct answer" behavior is currently reliable in single-turn terminal use. In the single-turn dcc-help invocation in the terminal, the model withholds the direct solution and instead guides the student.
Trained on a specific course/forum corpus, so its responses reflect the conventions, content, and any biases of that data and may be out of date relative to current course materials.
Like any LLM, it can produce incorrect or hallucinated explanations; student understanding should not depend solely on its output.

This gemma4 model was trained 2x faster with Unsloth and Huggingface's TRL library.

Downloads last month: 76

Safetensors

Model size

33B params

Tensor type

BF16