Instructions to use LLMWildling/gemma-4-80b-a8b-coder with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use LLMWildling/gemma-4-80b-a8b-coder with Transformers:

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

pipe = pipeline("text-generation", model="LLMWildling/gemma-4-80b-a8b-coder")
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, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("LLMWildling/gemma-4-80b-a8b-coder")
model = AutoModelForMultimodalLM.from_pretrained("LLMWildling/gemma-4-80b-a8b-coder", device_map="auto")
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 LLMWildling/gemma-4-80b-a8b-coder with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "LLMWildling/gemma-4-80b-a8b-coder"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "LLMWildling/gemma-4-80b-a8b-coder",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/LLMWildling/gemma-4-80b-a8b-coder

SGLang

How to use LLMWildling/gemma-4-80b-a8b-coder 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 "LLMWildling/gemma-4-80b-a8b-coder" \
    --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": "LLMWildling/gemma-4-80b-a8b-coder",
		"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 "LLMWildling/gemma-4-80b-a8b-coder" \
        --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": "LLMWildling/gemma-4-80b-a8b-coder",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use LLMWildling/gemma-4-80b-a8b-coder with Docker Model Runner:
```
docker model run hf.co/LLMWildling/gemma-4-80b-a8b-coder
```

LLMWildling/gemma-4-80b-a8b-coder issue with chat completion

by jaminy77 - opened 11 days ago

Discussion

jaminy77

11 days ago

•

edited 11 days ago

Hi,

I started this model with vLLM using the exact command provided in the model card:

vllm serve /ailab/vllm/models/gemma-4-80b-a8b-coder --served-model-name vllm/doobee --host 0.0.0.0 --port 23333 --dtype bfloat16 --tensor-parallel-size 1 --enable-expert-parallel --max-model-len 200000 --gpu-memory-utilization 0.96 --trust-remote-code --reasoning-parser gemma4 --tool-call-parser gemma4 --enable-auto-tool-choice --default-chat-template-kwargs '{"enable_thinking": true}' --generation-config vllm --language-model-only --skip-mm-profiling --max-num-seqs 1 --max-num-batched-tokens 8192

vLLM starts successfully, but every response consists of gibberish text, for example:

_material_反馈-disk-s-s-s-0-vector-vector-s-ve-version-v-v-s-ic-ic-과-ve-v-s-bed-s-ota-ton-i-f-target--시기-s-0-vector-v-s-ve-version-v-v-s-ic-ic-target--4-s-held-s-led-1-s-educ-red-red-red-empty-s-1-s-side-as-s-bed-os-ota-ton-i-f-target--시기-s-0-vector-v-s-ve-version-v-s-ic-ic-target--4-s-held-led-1-s-educ-red-red-red-empty-s-1-side-s-led-1-1-0-vector-v-s-s-0-vector-od-od-s-1-er-er-s-s- Feedback-disk-s-s-s-0-vector-vector-s-ve-version-v-v-s-ic-ic-target--4-s-held-led-1-s-educ-red-red-red-empty--ss-s-s-side-s-bed-os-s-ota-ton-i- & l-l-l- 完-callback-callback-s-s-0-vector-v-s-s-ve-version-v-s-ic-ic-target--4-s-held-led-1s-educ-red-red-red-empty-ss-s-side-bed-os-s-dot-1-red-feedback-disk-s-s-0-vector-v-vector-s-ve-version-_s-ic-ic-target-***-4 Skheld-https-https-s-s-s-s-0-vector-v-s-s-ve-ผม-M-M-cov-array-array-array-1-1-1-s-1- 0-vector-v-s-s-1-l-1-0-vector-v-s-
https-replace-system-
_וח-array-array-1-s-s-s- * -array-array1-s-ve-com-com-array-111-v-v-1-core-Tiny-Tiny-1 *Review-review-111-configure-system-array-array1-s-ve-com-com-array110-👋-6-0_parameter

I have successfully run other Gemma models on the same environment without any issues, ie LLMWildling/gemma-4-opencoder-40b-a8b-nvfp4 and LLMWildling/gemma-4-opencoder-40b-a8b-nvfp4-v2

Troubleshooting already performed:

Re-downloaded the chat_template.jinja file.
Tested through LiteLLM Chat, GitHub CLI, and direct Python API calls.

Does anyone have an idea what might be causing this?

GPU: RTX 6000 Blackwell Max-Q 96GB
Tried on vLLM 0.27.1 and vLLM in vllm:latest container 0.26.1rc1.dev608+g99a10304d

Thanks in advance.

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment