Instructions to use anonymousML123/Llama-3.1-8B-Tulu10pct-SFT-MAHALS with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use anonymousML123/Llama-3.1-8B-Tulu10pct-SFT-MAHALS with Transformers:

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

pipe = pipeline("text-generation", model="anonymousML123/Llama-3.1-8B-Tulu10pct-SFT-MAHALS")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("anonymousML123/Llama-3.1-8B-Tulu10pct-SFT-MAHALS")
model = AutoModelForCausalLM.from_pretrained("anonymousML123/Llama-3.1-8B-Tulu10pct-SFT-MAHALS", device_map="auto")
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 anonymousML123/Llama-3.1-8B-Tulu10pct-SFT-MAHALS with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "anonymousML123/Llama-3.1-8B-Tulu10pct-SFT-MAHALS"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "anonymousML123/Llama-3.1-8B-Tulu10pct-SFT-MAHALS",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/anonymousML123/Llama-3.1-8B-Tulu10pct-SFT-MAHALS

SGLang

How to use anonymousML123/Llama-3.1-8B-Tulu10pct-SFT-MAHALS 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 "anonymousML123/Llama-3.1-8B-Tulu10pct-SFT-MAHALS" \
    --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": "anonymousML123/Llama-3.1-8B-Tulu10pct-SFT-MAHALS",
		"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 "anonymousML123/Llama-3.1-8B-Tulu10pct-SFT-MAHALS" \
        --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": "anonymousML123/Llama-3.1-8B-Tulu10pct-SFT-MAHALS",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use anonymousML123/Llama-3.1-8B-Tulu10pct-SFT-MAHALS with Docker Model Runner:
```
docker model run hf.co/anonymousML123/Llama-3.1-8B-Tulu10pct-SFT-MAHALS
```

Llama-3.1-8B-Tulu10pct-SFT-MAHALS

Supervised Fine-Tuned (SFT) Llama 3.1 8B model trained on 10% of the Tulu-3 SFT mixture for the MAHALS research project.

Model Details

Attribute	Value
Base Model	meta-llama/Llama-3.1-8B
Training Method	Supervised Fine-Tuning (SFT)
Dataset	allenai/tulu-3-sft-mixture (10%)
Framework	allenai/open-instruct
License	Llama 3.1 Community License

Inference Requirements

Precision	VRAM Required	Compatible GPUs
BF16/FP16	~20 GB	A100 40GB, RTX 4090/3090, A10
INT8	~10 GB	T4, RTX 3080
INT4	~6 GB	RTX 3060, consumer GPUs

Usage

Transformers

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model_id = "anonymousML123/Llama-3.1-8B-Tulu10pct-SFT-MAHALS"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto"
)

messages = [{"role": "user", "content": "What is machine learning?"}]

input_ids = tokenizer.apply_chat_template(messages, return_tensors="pt").to(model.device)
outputs = model.generate(input_ids, max_new_tokens=256)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

vLLM Serving

vllm serve anonymousML123/Llama-3.1-8B-Tulu10pct-SFT-MAHALS --max_model_len=4096

Chat Template

This model uses the Tulu chat template:

<|user|>
Your question here
<|assistant|>
Model response<|endoftext|>

Training Configuration

Parameter	Value
Learning Rate	5e-6
Effective Batch Size	128
Gradient Accumulation	16
Max Sequence Length	4096
Epochs	2
LR Schedule	Linear
Warmup Ratio	0.03
Optimizer	AdamW
Precision	BF16

Training Data

10% random sample (~94K examples) from allenai/tulu-3-sft-mixture, which includes:

FLAN v2
Open Assistant
ShareGPT
Code instructions
Math instructions

Intended Use

This model is intended for research on multi-agent alignment and instruction following. It is part of the MAHALS (Multi-Agent Hierarchical Alignment) research project.

Limitations

Trained on 10% of data (reduced capability vs full Tulu-3)
English only
May exhibit biases present in training data
Not suitable for production without further evaluation

Citation

@misc{mahals2026,
  title={MAHALS: Multi-Agent Hierarchical Alignment},
  author={Anonymous},
  year={2026},
  note={Under review}
}

Acknowledgments

Built using AllenAI's open-instruct framework.

Downloads last month: 10

Model tree for anonymousML123/Llama-3.1-8B-Tulu10pct-SFT-MAHALS

Base model

meta-llama/Llama-3.1-8B

Finetuned

(1479)

this model

Finetunes

1 model

anonymousML123
/

Llama-3.1-8B-Tulu10pct-SFT-MAHALS