Instructions to use duoyuncloud/zsh-autocomplete-lora with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use duoyuncloud/zsh-autocomplete-lora with PEFT:

from peft import PeftModel
from transformers import AutoModelForCausalLM

base_model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-0.5B-Instruct")
model = PeftModel.from_pretrained(base_model, "duoyuncloud/zsh-autocomplete-lora")

Transformers

How to use duoyuncloud/zsh-autocomplete-lora with Transformers:

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

pipe = pipeline("text-generation", model="duoyuncloud/zsh-autocomplete-lora")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("duoyuncloud/zsh-autocomplete-lora", dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use duoyuncloud/zsh-autocomplete-lora with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/duoyuncloud/zsh-autocomplete-lora

SGLang

How to use duoyuncloud/zsh-autocomplete-lora 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 "duoyuncloud/zsh-autocomplete-lora" \
    --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": "duoyuncloud/zsh-autocomplete-lora",
		"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 "duoyuncloud/zsh-autocomplete-lora" \
        --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": "duoyuncloud/zsh-autocomplete-lora",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use duoyuncloud/zsh-autocomplete-lora with Docker Model Runner:
```
docker model run hf.co/duoyuncloud/zsh-autocomplete-lora
```

zsh-autocomplete-lora

LoRA adapter for local Zsh CLI command completion, used by the duoyuncloud/zsh-llm-cli-autocomplete-tool daemon runtime.

Model Details

Adapter type: PEFT LoRA (task_type=CAUSAL_LM)
Base model: Qwen/Qwen2.5-0.5B-Instruct
Primary use case: inline shell command completion in Zsh
Language: shell command text (plus lightweight natural language in prompts)

LoRA Configuration

From adapter_config.json:

r: 16
lora_alpha: 32
lora_dropout: 0.05
target_modules: q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj
inference_mode: true

Intended Use

Low-latency local command completion for terminal workflows
Context-aware completion with shell history and lightweight repository context from host app
Smart commit command suggestion (handled by host app logic + this adapter)

Out-of-Scope Use

Safety-critical automation without user review
General-purpose factual Q&A or long-form generation
Execution of generated commands without human confirmation

Quick Usage

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

base_id = "Qwen/Qwen2.5-0.5B-Instruct"
adapter_id = "duoyuncloud/zsh-autocomplete-lora"

tokenizer = AutoTokenizer.from_pretrained(base_id, trust_remote_code=True)
base = AutoModelForCausalLM.from_pretrained(base_id, trust_remote_code=True)
model = PeftModel.from_pretrained(base, adapter_id)

# Optional: merge for faster standalone inference
merged = model.merge_and_unload()

For this project, installation and daemon wiring are handled by:

install.sh in the project repository
python -m model_completer.daemon runtime entrypoint

Training Notes

This adapter is trained with SFT-style shell completion data (instruction / input / output) and optimized for local interactive inference.

Limitations

May suggest incorrect or unsafe commands in ambiguous contexts
Performance depends on user history/context quality from host integration
Should be used with explicit user acceptance before command execution

Safety

Keep human-in-the-loop confirmation for command execution
Apply allow/deny checks in host app for risky flags/operations
Avoid using generated output as direct automation input

Model tree for duoyuncloud/zsh-autocomplete-lora

Base model

Qwen/Qwen2.5-0.5B

Finetuned

Qwen/Qwen2.5-0.5B-Instruct

Adapter

(579)

this model

duoyuncloud
/

zsh-autocomplete-lora