metadata

license: llama3.1
base_model: meta-llama/Meta-Llama-3.1-8B
tags:
  - axolotl
  - generated_from_trainer
datasets:
  - Magpie-Align/Magpie-Llama-3.1-Pro-500K-Filtered
  - Magpie-Align/Magpie-Reasoning-150K
model-index:
  - name: Llama-3.1-8B-Magpie-SFT-650KR
    results: []

🐦 Llama-3.1-8B-Magpie-Align-SFT-v0.2

Project Web: https://magpie-align.github.io/

Arxiv Technical Report: https://arxiv.org/abs/2406.08464

Codes: https://github.com/magpie-align/magpie

Abstract

Click Here

High-quality instruction data is critical for aligning large language models (LLMs). Although some models, such as Llama-3-Instruct, have open weights, their alignment data remain private, which hinders the democratization of AI. High human labor costs and a limited, predefined scope for prompting prevent existing open-source data creation methods from scaling effectively, potentially limiting the diversity and quality of public alignment datasets. Is it possible to synthesize high-quality instruction data at scale by extracting it directly from an aligned LLM? We present a self-synthesis method for generating large-scale alignment data named Magpie. Our key observation is that aligned LLMs like Llama-3-Instruct can generate a user query when we input only the left-side templates up to the position reserved for user messages, thanks to their auto-regressive nature. We use this method to prompt Llama-3-Instruct and generate 4 million instructions along with their corresponding responses. We perform a comprehensive analysis of the extracted data and select 300K high-quality instances. To compare Magpie data with other public instruction datasets, we fine-tune Llama-3-8B-Base with each dataset and evaluate the performance of the fine-tuned models. Our results indicate that in some tasks, models fine-tuned with Magpie perform comparably to the official Llama-3-8B-Instruct, despite the latter being enhanced with 10 million data points through supervised fine-tuning (SFT) and subsequent feedback learning. We also show that using Magpie solely for SFT can surpass the performance of previous public datasets utilized for both SFT and preference optimization, such as direct preference optimization with UltraFeedback. This advantage is evident on alignment benchmarks such as AlpacaEval, ArenaHard, and WildBench.

About This Model

This model is a fine-tuned version of meta-llama/Meta-Llama-3.1-8B on

It achieves performance comparable with the official Llama-3.1-8B-Instruct Model with SFT only!

Alpaca Eval 2 (GPT-4-Turbo-1106): 20.66 (LC), 22.26 (WR)
Arena Hard: 22.2

Other Information

License: Please follow Meta Llama 3.1 Community License.

Conversation Template: Please use Llama 3 official chat template for the best performance.

Citation

If you find the model, data, or code useful, please cite our paper:

@article{xu2024magpie,
    title={Magpie: Alignment Data Synthesis from Scratch by Prompting Aligned LLMs with Nothing}, 
    author={Zhangchen Xu and Fengqing Jiang and Luyao Niu and Yuntian Deng and Radha Poovendran and Yejin Choi and Bill Yuchen Lin},
    year={2024},
    eprint={2406.08464},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 2e-05
train_batch_size: 1
eval_batch_size: 1
seed: 42
distributed_type: multi-GPU
num_devices: 4
gradient_accumulation_steps: 32
total_train_batch_size: 128
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: 65
num_epochs: 2

Training results

Training Loss	Epoch	Step	Validation Loss
0.6921	0.0029	1	0.7830
0.4187	0.1998	69	0.4135
0.3744	0.3997	138	0.3695
0.36	0.5995	207	0.3549
0.3603	0.7993	276	0.3459
0.3517	0.9992	345	0.3407
0.3064	1.1881	414	0.3392
0.3149	1.3879	483	0.3378
0.304	1.5877	552	0.3372
0.3059	1.7876	621	0.3370
0.323	1.9874	690	0.3370

Framework versions

Transformers 4.43.3
Pytorch 2.4.0+cu121
Datasets 2.19.1
Tokenizers 0.19.1

See axolotl config

axolotl version: 0.4.1


base_model: meta-llama/Meta-Llama-3.1-8B
model_type: LlamaForCausalLM
tokenizer_type: AutoTokenizer

load_in_8bit: false
load_in_4bit: false
strict: false

datasets:
  - path: Magpie-Align/Magpie-Reasoning-150K
    type: sharegpt
    conversation: llama3
  - path: Magpie-Align/Magpie-Llama-3.1-Pro-500K-Filtered
    type: sharegpt
    conversation: llama3
dataset_prepared_path: last_run_prepared
val_set_size: 0.001
output_dir: saves/llama3-mix-text-650KR

sequence_len: 8192
sample_packing: true
eval_sample_packing: false
pad_to_sequence_len: true

wandb_project: SynDa
wandb_entity:
wandb_watch:
wandb_name: Llama-3.1-8B-Magpie-Align-SFT-v0.2
wandb_log_model:
hub_model_id: Magpie-Align/Llama-3.1-8B-Magpie-Align-SFT-v0.2

gradient_accumulation_steps: 32
micro_batch_size: 1
num_epochs: 2
optimizer: paged_adamw_8bit
lr_scheduler: cosine
learning_rate: 2e-5

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_ratio: 0.1
evals_per_epoch: 5
eval_table_size:
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
  pad_token: <|end_of_text|>