Instructions to use torenartificialintelligence/toren-500m-instruct with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use torenartificialintelligence/toren-500m-instruct with Transformers:

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

pipe = pipeline("text-generation", model="torenartificialintelligence/toren-500m-instruct")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("torenartificialintelligence/toren-500m-instruct", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use torenartificialintelligence/toren-500m-instruct with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "torenartificialintelligence/toren-500m-instruct"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "torenartificialintelligence/toren-500m-instruct",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/torenartificialintelligence/toren-500m-instruct

SGLang

How to use torenartificialintelligence/toren-500m-instruct 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 "torenartificialintelligence/toren-500m-instruct" \
    --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": "torenartificialintelligence/toren-500m-instruct",
		"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 "torenartificialintelligence/toren-500m-instruct" \
        --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": "torenartificialintelligence/toren-500m-instruct",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use torenartificialintelligence/toren-500m-instruct with Docker Model Runner:
```
docker model run hf.co/torenartificialintelligence/toren-500m-instruct
```

Model Card for Toren-500M-Instruct

Toren-500M-Instruct is a compact, instruction-tuned language model released by Toren Artificial Intelligence, weighing in at only around 127 MB. It's built to punch above its size — handling simple coding questions, geography, general advice, and everyday conversation — while staying small enough to run on modest hardware.

Model Details

Model Description

Toren-500M-Instruct is a low-parameter, low-storage model, fine-tuned to be as capable as possible for everyday tasks despite its small footprint. It isn't intended to match larger models on heavy coding or advanced reasoning, but it aims to close the gap on general knowledge, conversation, and lightweight assistance.

This is the model card for a 🤗 transformers model pushed to the Hub.

Developed by: Toren Artificial Intelligence Funded by: Self-funded (no external funding) Shared by: Toren Artificial Intelligence Model type: LoRA adapter (instruction-tuned) Language(s) (NLP): English License: Apache 2.0 Finetuned from model: Qwen/Qwen2.5-0.5B-Instruct Model Sources Repository: [More Information Needed] Paper: Not applicable Demo: [More Information Needed] Uses

Direct Use

Toren-500M-Instruct can be used directly for general-purpose chat: answering everyday questions, explaining geography and general knowledge topics, offering advice, writing and rewriting text, summarizing short passages, and helping with simple coding questions.

Downstream Use

The model is well suited to being embedded in lightweight applications — chatbots, Discord bots, desktop assistants, or educational tools — where a small footprint and fast inference matter more than top-tier reasoning ability.

Out-of-Scope Use

Toren-500M-Instruct should not be used as an authoritative source for medical, legal, or financial decisions, for tasks requiring guaranteed factual accuracy without human review, or for complex multi-step reasoning, advanced software engineering, or rigorous mathematics — its small size isn't built for that.

Bias, Risks, and Limitations

As a small model, Toren-500M-Instruct can produce incorrect or inconsistent information, especially on niche topics, and may struggle with multi-step or abstract reasoning. It has not been formally evaluated for bias or fairness, so outputs on sensitive topics should be reviewed critically rather than taken at face value.

Recommendations

Users (both direct and downstream) should be made aware of the model's risks, biases, and limitations above. For anything high-stakes, outputs should be verified by a human before being relied upon.

How to Get Started with the Model

Use the code below to get started with the model.

python from transformers import AutoTokenizer, AutoModelForCausalLM

model_id = "torenartificialintelligence/Toren-500M-Instruct"

tokenizer = AutoTokenizer.from_pretrained(model_id) model = AutoModelForCausalLM.from_pretrained(model_id)

messages = [ {"role": "user", "content": "What is the capital of France?"} ]

inputs = tokenizer.apply_chat_template( messages, return_tensors="pt", add_generation_prompt=True )

outputs = model.generate( inputs, max_new_tokens=100 )

print(tokenizer.decode(outputs[0], skip_special_tokens=True)) Training Details Training Data

[More Information Needed]

Training Procedure Preprocessing

[More Information Needed]

Training Hyperparameters Training regime: LoRA / PEFT instruction tuning — [More Information Needed] Speeds, Sizes, Times Model size: ~127 MB Parameters: ~500M [More Information Needed] Evaluation Testing Data, Factors & Metrics Testing Data

[More Information Needed]

Factors

[More Information Needed]

Metrics

[More Information Needed]

Results

[More Information Needed]

Summary

Formal benchmark results are not yet available. Planned evaluation areas include general knowledge, instruction following, basic mathematics, coding, reasoning, and conversational quality.

Model Examination

[More Information Needed]

Environmental Impact

Carbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).

Hardware Type: [More Information Needed] Hours used: [More Information Needed] Cloud Provider: [More Information Needed] Compute Region: [More Information Needed] Carbon Emitted: [More Information Needed] Technical Specifications Model Architecture and Objective

Decoder-only Transformer, fine-tuned from Qwen2.5-0.5B-Instruct via LoRA for instruction-following and conversational objectives.

Compute Infrastructure

[More Information Needed]

Hardware

[More Information Needed]

Software

🤗 Transformers, PEFT

Citation

BibTeX:

[More Information Needed]

APA:

[More Information Needed]

Glossary LoRA (Low-Rank Adaptation): A fine-tuning method that trains small additional weight matrices instead of updating the full model, keeping training cheap and the adapter file small. More Information

[More Information Needed]

Model Card Authors

Toren Artificial Intelligence

Model Card Contact

[More Information Needed]

Downloads last month: -; Downloads are not tracked for this model. How to track

Model tree for torenartificialintelligence/toren-500m-instruct

Base model

Qwen/Qwen2.5-0.5B

Finetuned

Qwen/Qwen2.5-0.5B-Instruct

Adapter

(726)

this model