Instructions to use prithivMLmods/Tulu-MathLingo-8B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use prithivMLmods/Tulu-MathLingo-8B with Transformers:

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

pipe = pipeline("text-generation", model="prithivMLmods/Tulu-MathLingo-8B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("prithivMLmods/Tulu-MathLingo-8B")
model = AutoModelForCausalLM.from_pretrained("prithivMLmods/Tulu-MathLingo-8B")
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

vLLM

How to use prithivMLmods/Tulu-MathLingo-8B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "prithivMLmods/Tulu-MathLingo-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": "prithivMLmods/Tulu-MathLingo-8B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/prithivMLmods/Tulu-MathLingo-8B

SGLang

How to use prithivMLmods/Tulu-MathLingo-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 "prithivMLmods/Tulu-MathLingo-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": "prithivMLmods/Tulu-MathLingo-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 "prithivMLmods/Tulu-MathLingo-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": "prithivMLmods/Tulu-MathLingo-8B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use prithivMLmods/Tulu-MathLingo-8B with Docker Model Runner:
```
docker model run hf.co/prithivMLmods/Tulu-MathLingo-8B
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Tulu-MathLingo-8B Model Files

The Tulu-MathLingo-8B model is a fine-tuned version of meta-llama/Llama-3.1-8B, optimized for solving mathematical word problems and reasoning tasks in English. The model integrates advanced language understanding and reasoning capabilities with a focus on providing solutions to math-related queries.

File Name	Size	Description	Upload Status
`.gitattributes`	1.57 kB	Configures LFS tracking for large files.	Updated
`README.md`	292 Bytes	Basic details about the uploaded model.	Updated
`config.json`	988 Bytes	Contains model architecture and metadata.	Uploaded
`generation_config.json`	241 Bytes	Parameters for text generation (e.g., length, temperature).	Uploaded
`model-00001-of-00004.safetensors`	4.98 GB	Part 1 of model weights.	Uploaded (LFS)
`model-00002-of-00004.safetensors`	5 GB	Part 2 of model weights.	Uploaded (LFS)
`model-00003-of-00004.safetensors`	4.92 GB	Part 3 of model weights.	Uploaded (LFS)
`model-00004-of-00004.safetensors`	1.17 GB	Part 4 of model weights.	Uploaded (LFS)
`model.safetensors.index.json`	25.4 kB	Index file for multi-part model weights.	Uploaded
`special_tokens_map.json`	462 Bytes	Maps special tokens (e.g., `<PAD>`, `<EOS>`).	Uploaded
`tokenizer.json`	17.2 MB	Full tokenizer configuration.	Uploaded (LFS)
`tokenizer_config.json`	57.6 kB	Metadata for tokenizer usage.	Uploaded

Sample Solve

Key Features

Multilingual Math Reasoning:
- Designed for solving complex math problems in English and Tulu.
Text Generation:
- Generates detailed and contextually accurate text responses.
Fine-Tuned Specializations:
- Trained on the microsoft/orca-math-word-problems-200k dataset for word problem-solving.
Special Token Mapping:
- Configured to use tokens for specific functions such as <PAD> and <EOS> effectively.
Secure and Efficient Storage:
- Model weights are stored in the Safetensors format for secure and faster inference.
Large Parameter Size:
- 8.03 billion parameters enable handling complex queries and multi-turn conversations.

Training Details

Base Model: meta-llama/Llama-3.1-8B
Fine-Tuned:
- Through multiple stages: SFT (Supervised Fine-Tuning) and DPO (Direct Preference Optimization).
Dataset:
- Trained on 200k word problems from the Microsoft Orca Math Word Problems Dataset.
Model Size:
- 8.03B parameters, optimized for FP16 tensor type.

Applications

Mathematical Word Problems:
- Solve structured or unstructured math problems in natural language.
Conversational AI for Math:
- Engage users in interactive dialogues focused on math and logic reasoning.
Multilingual Support:
- Supports queries in Tulu and English, enhancing accessibility.
Education Tools:
- Useful in tutoring systems for math, helping students with problem-solving.

Usage

Loading the Model

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "prithivMLmods/Tulu-MathLingo-8B"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype="fp16")

Math Word Problem

query = "If a train travels 60 miles in 2 hours, what is its average speed?"
inputs = tokenizer(query, return_tensors="pt")
outputs = model.generate(**inputs, max_length=100)

response = tokenizer.decode(outputs[0], skip_special_tokens=True)
print("Answer:", response)

Performance Requirements

Hardware:
- Requires a GPU with at least 24GB VRAM for optimal performance due to model size and FP16 usage.
Optimization:
- Use mixed precision (fp16) for reduced memory footprint.
- Split inference across multiple GPUs if necessary.