---
license: cc-by-nc-4.0
tags:
- generated_from_trainer
base_model: lightblue/suzume-llama-3-8B-multilingual
model-index:
- name: workspace/llm_training/axolotl/llama3-multilingual-orpo/output_mitsu_top25_borda
results: []
---
# Suzume ORPO
[[Paper]](https://arxiv.org/abs/2405.18952) [[Dataset]](https://huggingface.co/datasets/lightblue/mitsu)
This is Suzume ORPO, an ORPO trained fine-tune of the [lightblue/suzume-llama-3-8B-multilingual](https://huggingface.co/lightblue/suzume-llama-3-8B-multilingual) model using our [lightblue/mitsu](https://huggingface.co/datasets/lightblue/mitsu) dataset.
We have trained several versions of this model using ORPO and so recommend that you use the best performing model from our tests, [lightblue/suzume-llama-3-8B-multilingual-orpo-borda-half](https://huggingface.co/lightblue/suzume-llama-3-8B-multilingual-orpo-borda-half).
Note that this model has a non-commerical license as we used the Command R and Command R+ models to generate our training data for this model ([lightblue/mitsu](https://huggingface.co/datasets/lightblue/mitsu)).
We are currently working on a developing a commerically usable model, so stay tuned for that!
# Model list
We have ORPO trained the following models using different proportions of the [lightblue/mitsu](https://huggingface.co/datasets/lightblue/mitsu) dataset:
* Trained on the top/bottom responses of all prompts in the dataset: [lightblue/suzume-llama-3-8B-multilingual-orpo-borda-full](https://huggingface.co/lightblue/suzume-llama-3-8B-multilingual-orpo-borda-full)
* Trained on the top/bottom responses of the prompts of the 75\% most consistently ranked responses in the dataset: [lightblue/suzume-llama-3-8B-multilingual-orpo-borda-top75](https://huggingface.co/lightblue/suzume-llama-3-8B-multilingual-orpo-borda-top75)
* Trained on the top/bottom responses of the prompts of the 50\% most consistently ranked responses in the dataset: [lightblue/suzume-llama-3-8B-multilingual-orpo-borda-half](https://huggingface.co/lightblue/suzume-llama-3-8B-multilingual-orpo-borda-half)
* Trained on the top/bottom responses of the prompts of the 25\% most consistently ranked responses in the dataset: [lightblue/suzume-llama-3-8B-multilingual-orpo-borda-top25](https://huggingface.co/lightblue/suzume-llama-3-8B-multilingual-orpo-borda-top25)
# Model results
We compare the MT-Bench scores across 6 languages for our 4 ORPO trained models, as well as some baselines:
* [meta-llama/Meta-Llama-3-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct) - The foundation model that our models are ultimately built upon
* [Nexusflow/Starling-LM-7B-beta](https://huggingface.co/Nexusflow/Starling-LM-7B-beta) - The highest performing open model on the Chatbot arena that is of a similar size to ours
* gpt-3.5-turbo - A fairly high quality (although not state-of-the-art) proprietary LLM
* [lightblue/suzume-llama-3-8B-multilingual](https://huggingface.co/lightblue/suzume-llama-3-8B-multilingual) - The base model which we train our ORPO finetunes from
| **MT-Bench language** | **meta-llama/Meta-Llama-3-8B-Instruct** | **Nexusflow/Starling-LM-7B-beta** | **gpt-3.5-turbo** | **lightblue/suzume-llama-3-8B-multilingual** | **lightblue/suzume-llama-3-8B-multilingual-orpo-borda-full** | **lightblue/suzume-llama-3-8B-multilingual-orpo-borda-top75** | **lightblue/suzume-llama-3-8B-multilingual-orpo-borda-half** | **lightblue/suzume-llama-3-8B-multilingual-orpo-borda-top25** |
|-----------------------|-----------------------------------------|-----------------------------------|-------------------|----------------------------------------------|--------------------------------------------------------------|---------------------------------------------------------------|--------------------------------------------------------------|---------------------------------------------------------------|
| **Chinese π¨π³** | NaN | 6.97 | 7.55 | 7.11 | 7.65 | **7.77** | 7.74 | 7.44 |
| **English πΊπΈ** | 7.98 | 7.92 | **8.26** | 7.73 | 7.98 | 7.94 | 7.98 | 8.22 |
| **French π«π·** | NaN | 7.29 | 7.74 | 7.66 | **7.84** | 7.46 | 7.78 | 7.81 |
| **German π©πͺ** | NaN | 6.99 | 7.68 | 7.26 | 7.28 | 7.64 | 7.7 | **7.71** |
| **Japanese π―π΅** | NaN | 6.22 | **7.84** | 6.56 | 7.2 | 7.12 | 7.34 | 7.04 |
| **Russian π·πΊ** | NaN | 8.28 | 7.94 | 8.19 | 8.3 | 8.74 | **8.94** | 8.81 |
We can see noticable improvement on most languages compared to the base model. We also find that our ORPO models achieve the highest score out of all the models we evaluated for a number of languages.
# Training data
We trained this model using the [lightblue/mitsu_full_borda](https://huggingface.co/datasets/lightblue/mitsu_full_borda) dataset.
# Training configuration
[
](https://github.com/OpenAccess-AI-Collective/axolotl)
See axolotl config
axolotl version: `0.4.0`
```yaml
base_model: lightblue/suzume-llama-3-8B-multilingual
model_type: LlamaForCausalLM
tokenizer_type: AutoTokenizer # PreTrainedTokenizerFast
load_in_8bit: false
load_in_4bit: false
strict: false
rl: orpo
orpo_alpha: 0.1
remove_unused_columns: false
chat_template: chatml
datasets:
- path: lightblue/mitsu_top25_borda
type: orpo.chat_template
conversation: llama-3
dataset_prepared_path: /workspace/llm_training/axolotl/llama3-multilingual-orpo/prepared_mitsu_top25_borda
val_set_size: 0.02
output_dir: /workspace/llm_training/axolotl/llama3-multilingual-orpo/output_mitsu_top25_borda
sequence_len: 8192
sample_packing: false
pad_to_sequence_len: true
use_wandb: true
wandb_project: axolotl
wandb_entity: peterd
wandb_name: mitsu_top25_borda
gradient_accumulation_steps: 8
micro_batch_size: 1
num_epochs: 1
optimizer: paged_adamw_8bit
lr_scheduler: cosine
learning_rate: 8e-6
train_on_inputs: false
group_by_length: false
bf16: auto
fp16:
tf32: false
gradient_checkpointing: true
gradient_checkpointing_kwargs:
use_reentrant: false
early_stopping_patience:
resume_from_checkpoint:
logging_steps: 1
xformers_attention:
flash_attention: true
warmup_steps: 10
evals_per_epoch: 20
eval_table_size:
saves_per_epoch: 1
debug:
deepspeed: /workspace/axolotl/deepspeed_configs/zero3_bf16.json
weight_decay: 0.0
special_tokens:
pad_token: <|end_of_text|>
```
# workspace/llm_training/axolotl/llama3-multilingual-orpo/output_mitsu_top25_borda
This model is a fine-tuned version of [lightblue/suzume-llama-3-8B-multilingual](https://huggingface.co/lightblue/suzume-llama-3-8B-multilingual) on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 0.0818
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 8e-06
- train_batch_size: 1
- eval_batch_size: 1
- seed: 42
- distributed_type: multi-GPU
- num_devices: 4
- gradient_accumulation_steps: 8
- total_train_batch_size: 32
- total_eval_batch_size: 4
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 10
- num_epochs: 1
### Training results
| Training Loss | Epoch | Step | Validation Loss |
|:-------------:|:-----:|:----:|:---------------:|
| 7.6328 | 0.05 | 1 | 7.7812 |
| 7.7158 | 0.1 | 2 | 7.2589 |
| 7.2588 | 0.15 | 3 | 4.0580 |
| 4.0068 | 0.19 | 4 | 2.4598 |
| 2.4438 | 0.24 | 5 | 0.6504 |
| 0.6586 | 0.29 | 6 | 0.1129 |
| 0.1235 | 0.34 | 7 | 0.1066 |
| 0.1273 | 0.39 | 8 | 0.1041 |
| 0.1076 | 0.44 | 9 | 0.0987 |
| 0.1009 | 0.48 | 10 | 0.0940 |
| 0.1172 | 0.53 | 11 | 0.0885 |
| 0.1016 | 0.58 | 12 | 0.0867 |
| 0.1088 | 0.63 | 13 | 0.0859 |
| 0.095 | 0.68 | 14 | 0.0846 |
| 0.1101 | 0.73 | 15 | 0.0839 |
| 0.0969 | 0.78 | 16 | 0.0832 |
| 0.0864 | 0.82 | 17 | 0.0825 |
| 0.0918 | 0.87 | 18 | 0.0821 |
| 0.0927 | 0.92 | 19 | 0.0819 |
| 0.0967 | 0.97 | 20 | 0.0818 |
### Framework versions
- Transformers 4.38.2
- Pytorch 2.2.1+cu121
- Datasets 2.18.0
- Tokenizers 0.15.0
# How to cite
```tex
@article{devine2024sure,
title={Are You Sure? Rank Them Again: Repeated Ranking For Better Preference Datasets},
author={Devine, Peter},
journal={arXiv preprint arXiv:2405.18952},
year={2024}
}
```
# Developer
Peter Devine - ([ptrdvn](https://huggingface.co/ptrdvn))