Instructions to use adikuma/mumble-cleanup with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use adikuma/mumble-cleanup with Transformers:

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

pipe = pipeline("text-generation", model="adikuma/mumble-cleanup")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForMultimodalLM

tokenizer = AutoTokenizer.from_pretrained("adikuma/mumble-cleanup")
model = AutoModelForMultimodalLM.from_pretrained("adikuma/mumble-cleanup")
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=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use adikuma/mumble-cleanup with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/adikuma/mumble-cleanup

SGLang

How to use adikuma/mumble-cleanup 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 "adikuma/mumble-cleanup" \
    --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": "adikuma/mumble-cleanup",
		"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 "adikuma/mumble-cleanup" \
        --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": "adikuma/mumble-cleanup",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use adikuma/mumble-cleanup with Docker Model Runner:
```
docker model run hf.co/adikuma/mumble-cleanup
```

mumble-cleanup / scripts /04_export.py

adikuma

initial upload: cleanup code and 688-pair seed dataset

fd0b01f verified 12 days ago

raw

history blame contribute delete

1.39 kB

	# merge lora into base, export fp32 onnx, quantize to int8.
	# writes runs/<run-id>/merged/, onnx/model.onnx, onnx/int8/model.onnx.

	import argparse
	from pathlib import Path

	from cleanup.config import load_train_config
	from cleanup.export.merge import merge_adapter
	from cleanup.export.quantize import quantize_int8
	from cleanup.export.to_onnx import export_onnx


	def main() -> None:
	parser = argparse.ArgumentParser()
	parser.add_argument("--config", default="configs/train.yaml")
	parser.add_argument("--runs-dir", default="runs")
	parser.add_argument("--run-id", required=True)
	parser.add_argument("--skip-int8", action="store_true")
	parser.add_argument("--skip-onnx", action="store_true", help="merge only, no onnx export")
	args = parser.parse_args()

	cfg = load_train_config(args.config)
	run_dir = Path(args.runs_dir) / args.run_id
	adapter_dir = run_dir / "model"
	merged_dir = run_dir / "merged"
	onnx_dir = run_dir / "onnx"
	int8_dir = onnx_dir / "int8"

	merge_adapter(cfg, adapter_dir, merged_dir)
	if args.skip_onnx:
	print("[export] skipping onnx per --skip-onnx")
	return
	fp32_onnx = export_onnx(merged_dir, onnx_dir)
	if not args.skip_int8:
	quantize_int8(fp32_onnx, int8_dir)
	print(f"next: make benchmark RUN_ID={args.run_id} (LOCAL cpu only)")


	if __name__ == "__main__":
	main()