Aobangaming/luna-2-70m

Model Card for Luna 2

Model Details

Luna 2 is a next generation text model designed for keeping context over long amounts of text. Luna 2 is generally intended for chatbot purposes.

• Model creator: AobanZ

Model Description

Luna utilizes FlashAttention and AdamW for performance and capability. Lightning is designed to provide quick, coherent outputs, improved with a larger size and weight.

• Developed by: AobanZ
• Model type: Transformer
• Language(s) (NLP): English
• License: MIT

Model Sources

• Repository: https://huggingface.co/Aobangaming/luna-2-70m

Uses

Luna is intended to be used for research, analysis and fine-tuning, general conversation, stories and other. It is not intended to be used for professional advice, real writing or any kind of heavy work as generated outputs may be incorrect.

Direct Use

Luna can be used directly for text generation, experimentation, and conversational interactions. Users can provide text prompts and generate responses using the model's built-in language modeling capabilities. Direct use is primarily intended for casual conversation, research and experimentation. Outputs may be incomplete, inaccurate, repetitive, or unrelated to the input, and should be evaluated before being used for other purposes.

Downstream Use

Luna may be fined-tuned for a AI Character, AI Agents, and chat models. However, please note that generated outputs may be corrupted and/or incorrect.

Out-of-Scope Use

Heavy Work may overload the model, which will cause corrupted outputs and/or misinformation. Please provide coherent, steady prompts as the model may not understand unfamiliar terms.

Bias, Risks, and Limitations

Lightning is designed to process english and conversational text ONLY and cannot be fined-tuned for any other uses(eg. Robotics)

Recommendations

We recommend users of Lightning to finetune the model on new text, and add necessary guardrails and precautions to prevent misuse.

How to Get Started with the Model

Use the code below to get started with the model.

import torch
from transformers import AutoModelForCausalLM
from tokenizers import Tokenizer
from huggingface_hub import hf_hub_download
import importlib.util

model_id = "Aobangaming/luna-2-70m"

path = hf_hub_download(model_id, "modeling_luna.py")
spec = importlib.util.spec_from_file_location("lightning", path)
module = importlib.util.module_from_spec(spec)
spec.loader.exec_module(module)

model = AutoModelForCausalLM.from_pretrained(
    model_id, trust_remote_code=True, ignore_mismatched_sizes=False
)
tokenizer = Tokenizer.from_file(
    hf_hub_download(model_id, "luna_tokenizer.json")
)

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = model.to(device)

# Generation(OPTIONAL)
    response = module.generate_text(
        model,
        tokenizer,
        prompt,
        max_len=100,
        device=device,
        top_k=40,
        top_p=0.6,
        penalty=1.2,
        temperature=0.8,
        chat_history=history
    )

Training Details

Training Data

Luna was trained a subset of the OASST dataset.

Training Procedure

Luna was trained on an RTX 3050 GPU, using FlashAttention/SDPA and MHA. The model was trained on a large dataset. It was not trained on fine-tuning datasets since memory issues.

Training Results

Epoch	Avg Loss	Perplexity
1	6.019369138	411.3190
2	4.677136674	107.4620
3	3.918399966	50.3199
4	3.308445575	27.3426
5	2.888957391	17.9746

Training Hyperparameters

Hyperparameter	Value	Comment
Precision	FP32
Optimizer	AdamW
Learning rate	4e-4
Batch size	32

Environmental Impact

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

• Hardware Type: RTX 3050 6GB
• Hours used: 1.5
• Cloud Provider: AobanLabs
• Compute Region: Asia
• Carbon Emitted: ~0.17 kg CO₂e

Technical Specifications

Model Architecture and Objective

Luna uses a 8-layer causal Transformer with 384-dimensional hidden states and 6 attention heads, inherited from previous models. The architecture is based on the base architecture of "Aoban Lightning", which uses pre-layer normalization, causal scaled dot-product attention, a 4× expansion GELU feed-forward network, sinusoidal positional encoding, and untied input/output embeddings.

Hyperparameter	Value	Comment
Layers	8
D_MODEL	384
Attention Heads	6
Vocabulary	~75003	w/ 200 Sequence length