🐦 Llama-3-8B-Magpie-Air-MT-SFT-v0.1-GGUF
This is quantized version of Magpie-Align/Llama-3-8B-Magpie-Air-MT-SFT-v0.1 created using llama.cpp
Model Description
Project Web: https://magpie-align.github.io/
Arxiv Technical Report: https://arxiv.org/abs/2406.08464
Codes: https://github.com/magpie-align/magpie
Abstract
About This Model
This model is a fine-tuned version of meta-llama/Meta-Llama-3-8B on Magpie-Align/Magpie-Air-MT-300K-v0.1 dataset.
It achieves performance comparable with the official Llama-3-8B-Instruct Model with SFT only!
- Alpaca Eval 2 (GPT-4-Turbo-1106): 22.98 (LC), 24.02 (WR)
- Alpaca Eval 2 (Llama-3-8B-Instruct): 49.63 (LC), 51.42 (WR)
- Arena Hard: 15.5
Other Information
License: Please follow Meta Llama 3 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:
@misc{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: 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: 100
- num_epochs: 2
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 0.7285 | 0.0007 | 1 | 0.7411 |
| 0.2863 | 0.3332 | 509 | 0.2875 |
| 0.2584 | 0.6664 | 1018 | 0.2501 |
| 0.2187 | 0.9996 | 1527 | 0.2282 |
| 0.1445 | 1.3130 | 2036 | 0.2246 |
| 0.1419 | 1.6462 | 2545 | 0.2211 |
| 0.1413 | 1.9794 | 3054 | 0.2210 |
Framework versions
- Transformers 4.40.2
- Pytorch 2.3.0+cu121
- Datasets 2.19.1
- Tokenizers 0.19.1
axolotl version: 0.4.0
base_model: meta-llama/Meta-Llama-3-8B
model_type: LlamaForCausalLM
tokenizer_type: AutoTokenizer
load_in_8bit: false
load_in_4bit: false
strict: false
datasets:
- path: SynDa/Llama-3-8B-SynDa-MultiRound-300K
type: sharegpt
conversation: llama3
dataset_prepared_path: last_run_prepared
val_set_size: 0.001
output_dir: ./out_Llama-3-70B-SynDa-300K-Multi-Round
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-3-70B-SynDa-300K-MR-2EP-FFT
wandb_log_model:
hub_model_id: SynDa/Llama-3-8B-SynDa-300K-MR
gradient_accumulation_steps: 8
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_steps: 100
evals_per_epoch: 3
eval_table_size:
saves_per_epoch: 3
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
pad_token: <|end_of_text|>