Instructions to use prithivMLmods/Acrux-500M-o1-Journey with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use prithivMLmods/Acrux-500M-o1-Journey with Transformers:

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

pipe = pipeline("text-generation", model="prithivMLmods/Acrux-500M-o1-Journey")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("prithivMLmods/Acrux-500M-o1-Journey")
model = AutoModelForCausalLM.from_pretrained("prithivMLmods/Acrux-500M-o1-Journey")
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 prithivMLmods/Acrux-500M-o1-Journey with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "prithivMLmods/Acrux-500M-o1-Journey"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "prithivMLmods/Acrux-500M-o1-Journey",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/prithivMLmods/Acrux-500M-o1-Journey

SGLang

How to use prithivMLmods/Acrux-500M-o1-Journey 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 "prithivMLmods/Acrux-500M-o1-Journey" \
    --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": "prithivMLmods/Acrux-500M-o1-Journey",
		"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 "prithivMLmods/Acrux-500M-o1-Journey" \
        --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": "prithivMLmods/Acrux-500M-o1-Journey",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use prithivMLmods/Acrux-500M-o1-Journey with Docker Model Runner:
```
docker model run hf.co/prithivMLmods/Acrux-500M-o1-Journey
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Acrux-500M-o1-Journey Model Files

The Acrux-500M-o1-Journey is a lightweight, instruction-tuned language model fine-tuned from the Qwen2.5-0.5B-Instruct base model. With a size of 500 million parameters, it is designed for cost-effective deployment and fast text generation while maintaining quality performance for instruction-following tasks.

File Name	Size	Description	Upload Status
`.gitattributes`	1.57 kB	Git attributes for managing LFS files.	Uploaded
`README.md`	195 Bytes	Model overview or documentation.	Updated
`added_tokens.json`	657 Bytes	Custom tokens for the tokenizer.	Uploaded
`config.json`	859 Bytes	Model configuration file.	Uploaded
`generation_config.json`	280 Bytes	Configuration for text generation.	Uploaded
`merges.txt`	1.82 MB	Merge rules for byte-pair encoding (BPE).	Uploaded
`pytorch_model.bin`	988 MB	Model weights (PyTorch format).	Uploaded (LFS)
`special_tokens_map.json`	644 Bytes	Mapping for special tokens.	Uploaded
`tokenizer.json`	11.4 MB	Full tokenizer configuration.	Uploaded (LFS)
`tokenizer_config.json`	7.73 kB	Additional tokenizer settings.	Uploaded
`vocab.json`	2.78 MB	Vocabulary for the tokenizer.	Uploaded

Key Features:

Compact Size with Efficient Performance:
The smaller parameter count (500M) ensures faster inference and reduced hardware requirements.
Instruction Optimization:
Fine-tuned to follow prompts effectively, making it suitable for interactive applications and prompt-based tasks.
Domain-Specific Training:
Trained on the GAIR/o1-journey dataset, providing tailored capabilities for specific use cases.

Training Details:

Base Model: Qwen2.5-0.5B-Instruct
Dataset Used for Fine-Tuning: GAIR/o1-journey
- A compact dataset focusing on instruction-driven generation with 1.42k samples.

Capabilities:

Instruction Following:
- Generates accurate and coherent responses to user instructions.
- Handles summarization, question-answering, and conversational tasks.
Fast Inference:
- Ideal for real-time applications due to reduced latency from its smaller size.
Interactive AI Development:
- Suitable for chatbots, virtual assistants, and instructional interfaces.

Usage Instructions:

Setup:
Download all model files, ensuring compatibility with the Hugging Face Transformers library.

Loading the Model:

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "prithivMLmods/Acrux-500M-o1-Journey"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

Sample Generate Text:

input_text = "Explain the concept of machine learning in simple terms."
inputs = tokenizer(input_text, return_tensors="pt")
outputs = model.generate(**inputs, max_length=100, temperature=0.7)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Optimize Generation:
Adjust parameters in generation_config.json for better control of output, such as:
- temperature for randomness.
- top_p for sampling diversity.
- max_length for output size.