Instructions to use thomasjvu/lisper-gemma4-e2b-audio-full with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use thomasjvu/lisper-gemma4-e2b-audio-full with Transformers:

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

pipe = pipeline("image-text-to-text", model="thomasjvu/lisper-gemma4-e2b-audio-full")
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("thomasjvu/lisper-gemma4-e2b-audio-full")
model = AutoModelForImageTextToText.from_pretrained("thomasjvu/lisper-gemma4-e2b-audio-full")
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

vLLM

How to use thomasjvu/lisper-gemma4-e2b-audio-full with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "thomasjvu/lisper-gemma4-e2b-audio-full"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "thomasjvu/lisper-gemma4-e2b-audio-full",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/thomasjvu/lisper-gemma4-e2b-audio-full

SGLang

How to use thomasjvu/lisper-gemma4-e2b-audio-full 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 "thomasjvu/lisper-gemma4-e2b-audio-full" \
    --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": "thomasjvu/lisper-gemma4-e2b-audio-full",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

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 "thomasjvu/lisper-gemma4-e2b-audio-full" \
        --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": "thomasjvu/lisper-gemma4-e2b-audio-full",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Unsloth Studio

How to use thomasjvu/lisper-gemma4-e2b-audio-full 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 thomasjvu/lisper-gemma4-e2b-audio-full 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 thomasjvu/lisper-gemma4-e2b-audio-full to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for thomasjvu/lisper-gemma4-e2b-audio-full to start chatting

Load model with FastModel

pip install unsloth
from unsloth import FastModel
model, tokenizer = FastModel.from_pretrained(
    model_name="thomasjvu/lisper-gemma4-e2b-audio-full",
    max_seq_length=2048,
)

Docker Model Runner
How to use thomasjvu/lisper-gemma4-e2b-audio-full with Docker Model Runner:
```
docker model run hf.co/thomasjvu/lisper-gemma4-e2b-audio-full
```

lisper-gemma4-e2b-audio-full / README.md

thomasjvu

Finalize Lisper E2B full model card

879ed7a verified 14 days ago

preview code

raw

history blame contribute delete

1.78 kB

	---
	library_name: transformers
	base_model: google/gemma-4-E2B-it
	license: gemma
	tags:
	- gemma-4
	- audio
	- merged
	- unsloth
	- speech
	- lisper
	---

	# Lisper Gemma 4 E2B Audio Full Checkpoint

	This is the merged Lisper checkpoint: `google/gemma-4-E2B-it` with the trained Lisper LoRA adapter folded into a standalone `safetensors` model.

	## Model Lineage

	- Base model: `google/gemma-4-E2B-it`
	- Training: Unsloth supervised fine-tuning with QLoRA / LoRA
	- Merge method: base + LoRA adapter merged into a 16-bit checkpoint
	- Weight file: `model.safetensors`
	- Training rows: `16,000`
	- Validation rows: `2,000`
	- Held-out test rows: `2,000`

	This is not a dense full-parameter fine-tune. It is a merged base+LoRA checkpoint for easier deployment.

	## Evaluation

	The release-quality evaluation is the v18 hybrid acoustic+Gemma path:

	- Held-out rows: `2,000`
	- Hard errors: `0`
	- Verdict: `pass`
	- Class match: `0.976`
	- Clear/non-clear match: `0.989`
	- Exact four-line format: `1.0`
	- Reason/cue/encouragement present: `1.0`

	The evaluated pipeline uses acoustic features for the lisp-class hint and Gemma for structured coaching text and tone. Do not interpret these metrics as a pure direct-Gemma raw-audio classification result.

	## Deployment

	- Browser/WebGPU demo: `thomasjvu/lisper-gemma4-e2b-audio-onnx-q4f16`
	- Server-side ZeroGPU fallback: `thomasjvu/lisper-zerogpu`
	- Adapter-only package: `thomasjvu/lisper-gemma4-e2b-audio-lora`

	Use the q4f16 ONNX/WebGPU package for browser demos. Use this merged checkpoint as the server-side correctness reference.

	## Limitations

	- The lisp dataset is synthetically generated from speaker-disjoint source speech.
	- This is a practice assistant, not a medical diagnosis tool or a replacement for a speech-language pathologist.