Instructions to use decisionslab/Dlab-852-Mini with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use decisionslab/Dlab-852-Mini with Transformers:

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

pipe = pipeline("text-generation", model="decisionslab/Dlab-852-Mini", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("decisionslab/Dlab-852-Mini", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("decisionslab/Dlab-852-Mini", trust_remote_code=True)
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]:]))

MLX

How to use decisionslab/Dlab-852-Mini with MLX:

# Make sure mlx-lm is installed
# pip install --upgrade mlx-lm

# Generate text with mlx-lm
from mlx_lm import load, generate

model, tokenizer = load("decisionslab/Dlab-852-Mini")

prompt = "Write a story about Einstein"
messages = [{"role": "user", "content": prompt}]
prompt = tokenizer.apply_chat_template(
    messages, add_generation_prompt=True
)

text = generate(model, tokenizer, prompt=prompt, verbose=True)

Notebooks
Google Colab
Kaggle
Local Apps
LM Studio

vLLM

How to use decisionslab/Dlab-852-Mini with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "decisionslab/Dlab-852-Mini"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "decisionslab/Dlab-852-Mini",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/decisionslab/Dlab-852-Mini

SGLang

How to use decisionslab/Dlab-852-Mini 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 "decisionslab/Dlab-852-Mini" \
    --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": "decisionslab/Dlab-852-Mini",
		"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 "decisionslab/Dlab-852-Mini" \
        --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": "decisionslab/Dlab-852-Mini",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

MLX LM

How to use decisionslab/Dlab-852-Mini with MLX LM:

Generate or start a chat session

# Install MLX LM
uv tool install mlx-lm
# Interactive chat REPL
mlx_lm.chat --model "decisionslab/Dlab-852-Mini"

Run an OpenAI-compatible server

# Install MLX LM
uv tool install mlx-lm
# Start the server
mlx_lm.server --model "decisionslab/Dlab-852-Mini"
# Calling the OpenAI-compatible server with curl
curl -X POST "http://localhost:8000/v1/chat/completions" \
   -H "Content-Type: application/json" \
   --data '{
     "model": "decisionslab/Dlab-852-Mini",
     "messages": [
       {"role": "user", "content": "Hello"}
     ]
   }'

Docker Model Runner
How to use decisionslab/Dlab-852-Mini with Docker Model Runner:
```
docker model run hf.co/decisionslab/Dlab-852-Mini
```

You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

decisionslab/DLab-852-Mini

The Model decisionslab/DLab-852-Mini is a fine-tuned version of microsoft/Phi-3.5-mini-instruct, optimized to replicate and align with the cultural nuances of Hong Kong. This model leverages a large, curated dataset of Hong Kong cultural values to enhance decision-making simulations, behavioral reasoning, and social interaction modeling, achieving performance levels that outperform the base model in key cultural alignment benchmarks.

Model Overview

Model Name: decisionslab/DLab-852-Mini
Base Model: microsoft/Phi-3.5-mini-instruct
Intended Use: Simulation of Hong Kong cultural behavior, decision-making processes, and social interactions.
Language(s): Primarily English

Model Description

decisionslab/DLab-852-Mini is designed to emulate the everyday life, conversations, and behavioral shifts of Hong Kongers by incorporating a dataset rich in local cultural values. Recognizing that cultural context significantly influences decision-making and social behavior, our fine-tuning process explicitly addresses the subjective nature of human behavior.

Key enhancements include:

Cultural Alignment: Fine-tuning on Hong Kong-specific datasets to capture local cultural nuances.
Benchmark Improvements:
- Individualism (IDV): 86% closer alignment to Hong Kong's cultural benchmark compared to the base Phi-3.5-mini.
- Uncertainty Avoidance (UAI): 84% better alignment versus Phi-3.5-mini.
- Indulgence (IVR): 74% improved mirroring of Hong Kong values.

These performance gains were validated using CD-Eval—a benchmark developed by researchers at Beijing Jiaotong University and Microsoft—demonstrating a significant cultural reasoning advantage.

Intended Use Cases

Cultural Simulation: Predicting and simulating behavioral trends and societal reactions in Hong Kong.
Policy and Decision Support: Enhancing simulation models for policy analysis and strategic decision-making.
Social Interaction Modeling: Tailoring conversational agents to interact authentically with Hong Kong users.
Academic and Market Research: Providing insights into culturally driven behavioral patterns.

Usage with mlx

Install the necessary package:

pip install mlx-lm

Load and generate a response with the model:

from mlx_lm import load, generate

model, tokenizer = load("decisionslab/DLab-852-Mini")

prompt = "你好, can you simulate a typical day in Hong Kong?"

if tokenizer.chat_template is not None:
    messages = [{"role": "user", "content": prompt}]
    prompt = tokenizer.apply_chat_template(messages, add_generation_prompt=True)

response = generate(model, tokenizer, prompt=prompt, verbose=True)
print(response)

Evaluation

The model's cultural alignment has been rigorously evaluated using CDEval:

Individualism (IDV): 86% closer alignment to Hong Kong cultural benchmarks.
Uncertainty Avoidance (UAI): 84% improvement over Phi-3.5-mini.
Indulgence (IVR): 74% better performance in mirroring local values.

These evaluations underscore the model’s effectiveness in capturing the subtleties of Hong Kong's cultural landscape.

License

All content in this repository is proprietary and confidential. The software and any associated documentation are the exclusive property of Decisions Lab. Unauthorized copying, distribution, modification, or use via any medium is strictly prohibited. Use of this software requires explicit permission from Decisions Lab.

Contact

For inquiries, collaborations, or feedback, please contact Decisions Lab via hello@decisionslab.com.

Downloads last month: -

Safetensors

Model size

0.6B params

Tensor type

F16

U32

MLX

Hardware compatibility

Quantized