Update README.md

README.md

@@ -1,6 +1,6 @@
 ---
 license: llama3.1
-base_model: Magpie-Align/Llama-3.1-8B-Magpie-SFT-650KR
+base_model: Magpie-Align/Llama-3.1-8B-Magpie-Align-SFT-v0.2
 tags:
 - alignment-handbook
 - trl
@@ -12,40 +12,57 @@ tags:
 datasets:
 - Magpie-Align/Llama-3.1-70B-PO-100K-armorm
 model-index:
-- name: Llama-3.1-8B-Magpie-SFT-650KR-Magpo-Armorm-3.1-70B-05
+- name: Llama-3.1-8B-Magpie-Align-v0.2
   results: []
+language:
+- en
 ---
 
-<!-- This model card has been generated automatically according to the information the Trainer had access to. You
-should probably proofread and complete it, then remove this comment. -->
+![Magpie](https://cdn-uploads.huggingface.co/production/uploads/653df1323479e9ebbe3eb6cc/FWWILXrAGNwWr52aghV0S.png)
+## 🔥 Chat with Magpie [Here](https://huggingface.co/spaces/flydust/Chat-with-Magpie)!
 
-# Llama-3.1-8B-Magpie-SFT-650KR-Magpo-Armorm-3.1-70B-05
+# 🐦 Llama-3-8B-Magpie-Align-v0.3
 
-This model is a fine-tuned version of [Magpie-Align/Llama-3.1-8B-Magpie-SFT-650KR](https://huggingface.co/Magpie-Align/Llama-3.1-8B-Magpie-SFT-650KR) on the Magpie-Align/Llama-3.1-70B-PO-100K-armorm dataset.
-It achieves the following results on the evaluation set:
-- Loss: 0.3328
-- Rewards/chosen: -4.8597
-- Rewards/rejected: -7.5767
-- Rewards/accuracies: 0.8900
-- Rewards/margins: 2.7170
-- Logps/rejected: -1107.7988
-- Logps/chosen: -829.9976
-- Logits/rejected: -0.8134
-- Logits/chosen: -0.8059
+Project Web: [https://magpie-align.github.io/](https://magpie-align.github.io/)
 
-## Model description
+Online Model Demo: [https://huggingface.co/spaces/flydust/Chat-with-Magpie](https://huggingface.co/spaces/flydust/Chat-with-Magpie)
 
-More information needed
+Arxiv Technical Report: [https://arxiv.org/abs/2406.08464](https://arxiv.org/abs/2406.08464)
 
-## Intended uses & limitations
+Codes: [https://github.com/magpie-align/magpie](https://github.com/magpie-align/magpie)
 
-More information needed
+## 🧐 About This Model
 
-## Training and evaluation data
+This model is an aligned version of [meta-llama/Meta-Llama-3.1-8B](https://huggingface.co/meta-llama/Meta-Llama-3.1-8B). We apply the following pipeline:
 
-More information needed
+We first perform SFT using:
+* [Magpie-Align/Magpie-Llama-3.1-Pro-500K-Filtered](https://huggingface.co/datasets/Magpie-Align/Magpie-Llama-3.1-Pro-500K-Filtered)
+* [Magpie-Align/Magpie-Reasoning-150K](https://huggingface.co/datasets/Magpie-Align/Magpie-Reasoning-150K)
+* **SFT Model Checkpoint:** [Magpie-Align/Llama-3.1-8B-Magpie-Align-SFT-v0.2](https://huggingface.co/Magpie-Align/Llama-3.1-8B-Magpie-Align-SFT-v0.2)
 
-## Training procedure
+We then perform DPO on the [Magpie-Align/Llama-3.1-70B-PO-100K-armorm](https://huggingface.co/datasets/Magpie-Align/Llama-3.1-70B-PO-100K-armorm) dataset.
+
+The overall performance is much better than the official Llama-3.1-8B-Instruct Model! 
+
+- **Alpaca Eval 2 (vs GPT-4-Turbo-1106): 46.68 (LC), 53.42 (WR)**
+- **Arena Hard: 43.2**
+
+
+## 👀 Other Information
+
+**License**: Please follow [Meta Llama 3.1 Community License](https://github.com/meta-llama/llama-models/blob/main/models/llama3_1/LICENSE).
+
+**Conversation Template**: Please use Llama 3 **official chat template** for the best performance.
+
+**How to use it?** Please check the official [Llama 3.1 repository](https://huggingface.co/meta-llama/Meta-Llama-3.1-8B-Instruct#how-to-use) for detailed instructions. Simply replace the original `model_id` with this model id.
+
+The detailed training pipeline is as follows.
+
+## Stage 1: Supervised Fine-tuning
+
+We use [Axolotl](https://github.com/OpenAccess-AI-Collective/axolotl) for SFT. Please refer to the model card of [SFT checkpoint](https://huggingface.co/Magpie-Align/Llama-3.1-8B-Magpie-Align-SFT-v0.2) for detailed configurations.
+
+## Stage 2: Direct Preference Optimization
 
 ### Training hyperparameters
 
@@ -76,10 +93,93 @@ The following hyperparameters were used during training:
 | 0.3342        | 0.7837 | 600  | 0.3354          | -4.7041        | -7.3342          | 0.8920             | 2.6301          | -1083.5503     | -814.4402    | -0.8081         | -0.7999       |
 | 0.3251        | 0.9144 | 700  | 0.3335          | -4.8366        | -7.5394          | 0.8880             | 2.7028          | -1104.0730     | -827.6954    | -0.8119         | -0.8042       |
 
-
 ### Framework versions
 
 - Transformers 4.43.3
 - Pytorch 2.4.0+cu121
 - Datasets 2.20.0
 - Tokenizers 0.19.1
+
+<details><summary>See alignment handbook config</summary>
+
+```yaml
+# Customized Configs
+model_name_or_path: Magpie-Align/Llama-3.1-8B-Magpie-Align-SFT-v0.2
+hub_model_id: Magpie-Align/Llama-3.1-8B-Magpie-Align-v0.2
+output_dir: /data/zhangchen_xu/alignment_handbook_out/Llama-3.1-8B-Magpie-Align-v0.2
+run_name: Llama-3.1-8B-Magpie-Align-v0.2
+
+dataset_mixer:
+  Magpie-Align/Llama-3.1-70B-PO-100K-armorm: 1.0
+dataset_splits:
+- train
+- test
+preprocessing_num_workers: 64
+
+# DPOTrainer arguments
+bf16: true
+beta: 0.01
+learning_rate: 0.5e-6
+gradient_accumulation_steps: 16
+per_device_train_batch_size: 2
+per_device_eval_batch_size: 4
+num_train_epochs: 1
+max_length: 2048
+max_prompt_length: 1800
+warmup_ratio: 0.1
+logging_steps: 1
+lr_scheduler_type: cosine
+optim: adamw_torch
+
+torch_dtype: null
+use_flash_attention_2: true
+do_eval: true
+evaluation_strategy: steps
+eval_steps: 100
+gradient_checkpointing: true
+gradient_checkpointing_kwargs:
+  use_reentrant: False
+log_level: info
+push_to_hub: true
+save_strategy: "steps"
+save_steps: 100
+save_total_limit: 1
+seed: 42
+report_to:
+- wandb
+```
+</details><be>
+
+## Paper Abstract
+
+<details><summary>Click Here</summary>
+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.
+</details><be>
+
+## 📚 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}
+}
+```
+
+Please also cite the reward model for creating preference datasets:
+
+ArmoRM paper:
+```
+@article{wang2024interpretable,
+  title={Interpretable Preferences via Multi-Objective Reward Modeling and Mixture-of-Experts},
+  author={Wang, Haoxiang and Xiong, Wei and Xie, Tengyang and Zhao, Han and Zhang, Tong},
+  journal={arXiv preprint arXiv:2406.12845},
+  year={2024}
+}
+```
+
+**Questions?** Please contact [Zhangchen](https://zhangchenxu.com/) by email.