Instructions to use Alphaplasti/ToneBridge-MiniCPM4.1-8B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Alphaplasti/ToneBridge-MiniCPM4.1-8B with Transformers:

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

pipe = pipeline("text-generation", model="Alphaplasti/ToneBridge-MiniCPM4.1-8B", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("Alphaplasti/ToneBridge-MiniCPM4.1-8B", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Alphaplasti/ToneBridge-MiniCPM4.1-8B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Alphaplasti/ToneBridge-MiniCPM4.1-8B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Alphaplasti/ToneBridge-MiniCPM4.1-8B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Alphaplasti/ToneBridge-MiniCPM4.1-8B

SGLang

How to use Alphaplasti/ToneBridge-MiniCPM4.1-8B 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 "Alphaplasti/ToneBridge-MiniCPM4.1-8B" \
    --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": "Alphaplasti/ToneBridge-MiniCPM4.1-8B",
		"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 "Alphaplasti/ToneBridge-MiniCPM4.1-8B" \
        --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": "Alphaplasti/ToneBridge-MiniCPM4.1-8B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Alphaplasti/ToneBridge-MiniCPM4.1-8B with Docker Model Runner:
```
docker model run hf.co/Alphaplasti/ToneBridge-MiniCPM4.1-8B
```

ToneBridge-MiniCPM4.1-8B / training_data /README.md

Alphaplasti

Update model card and training data documentation

6204a10 verified about 2 months ago

preview code

Raw

History Blame Contribute Delete

1.38 kB

Training Data

This folder documents the task-specific datasets used to fine-tune ToneBridge-MiniCPM4.1-8B.

The files are JSONL files in chat prompt/completion format.

Files

File	Rows	Purpose
`hsk12_english_order_train.jsonl`	1,800	Training split for HSK 1/2-style sentences influenced by English word order.
`hsk12_english_order_valid.jsonl`	200	Validation split for the HSK 1/2 word-order task.
`context_tone_hsk3_train.jsonl`	4,500	Training split for HSK 1-3 context and tone adaptation.
`context_tone_hsk3_valid.jsonl`	500	Validation split for the HSK 1-3 context/tone task.

Dataset Format

Each row follows this structure:

{
  "prompt": [
    {
      "role": "system",
      "content": "你是中文语境校对助手。只输出更合适的句子，不要解释。"
    },
    {
      "role": "user",
      "content": "上下文：...\n原句：...\n任务：请根据上下文把原句改成更合适的中文。/no_think"
    }
  ],
  "completion": [
    {
      "role": "assistant",
      "content": "..."
    }
  ]
}

Notes

The data is synthetic and task-specific.
It is intended for beginner Mandarin correction and tone adaptation.
The dataset was generated to reduce exact duplicate source/correction pairs.
The data should not be treated as a general Chinese benchmark.