一个在多种公开数据集和合 datasets 上使用 DiscoPOP 进行微调的 7B 参数类似 GPT 模型
21 Pulls 更新于 2 个月前
更新于 2 个月前
2 个月前
a6c0012a5f2f · 5.7GB
README
DiscoPOP-zephyr-7b-gemma
此模型是 HuggingFaceH4/zephyr-7b-gemma-sft-v0.1 在 argilla/dpo-mix-7k 数据集上微调的版本。
此模型来自论文“Discovering Preference Optimization Algorithms with and for Large Language Models”。
在此处阅读有关它的博客文章!
在此处查看生成它的代码库: https://github.com/SakanaAI/DiscoPOP
模型描述
此模型在训练上与HuggingFaceH4/zephyr-7b-gemma-v0.1 完全相同,除了不使用直接偏好优化(DPO),而是使用 DiscoPOP。
DiscoPOP 是我们发现的偏好优化算法,其定义如下
def log_ratio_modulated_loss(
self,
policy_chosen_logps: torch.FloatTensor,
policy_rejected_logps: torch.FloatTensor,
reference_chosen_logps: torch.FloatTensor,
reference_rejected_logps: torch.FloatTensor,
) -> torch.FloatTensor:
pi_logratios = policy_chosen_logps - policy_rejected_logps
ref_logratios = reference_chosen_logps - reference_rejected_logps
logits = pi_logratios - ref_logratios
# Modulate the mixing coefficient based on the log ratio magnitudes
log_ratio_modulation = torch.sigmoid(logits)
logistic_component = -F.logsigmoid(self.beta * logits)
exp_component = torch.exp(-self.beta * logits)
# Blend between logistic and exponential component based on log ratio modulation
losses = logistic_component * (1 - log_ratio_modulation) + exp_component * log_ratio_modulation
return losses
训练超参数
以下超参数在训练期间使用
- learning_rate: 5e-07
- train_batch_size: 2
- eval_batch_size: 4
- seed: 42
- distributed_type: multi-GPU
- num_devices: 8
- gradient_accumulation_steps: 8
- total_train_batch_size: 128
- total_eval_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 2
框架版本
- Transformers 4.40.1
- Pytorch 2.1.2+cu121
- DataSets 2.19.0
- Tokenizers 0.19.1
Zephyr 7B Gemma 的模型卡
Zephyr 是一系列旨在充当有用助手的语言模型的系列模型。Zephyr 7B Gemma 是该系列中的第三个模型,是使用直接偏好优化(DPO)训练的公开可用和合成数据集混合的 google/gemma-7b
的微调版本。您可以通过在 Alignment Handbook 中提供的食谱来重现该模型的训练。
模型描述
- 模型类型: 在公开可用和合成数据集混合上微调的类似 GPT 的 7B 参数模型。
- 语言(NLP): 主要为英语
- 许可: Gemma 使用条款
- 微调自模型: google/gemma-7b
模型来源
- 仓库: https://github.com/huggingface/alignment-handbook
- 演示: https://hugging-face.cn/spaces/HuggingFaceH4/zephyr-7b-gemma-chat
性能
模型 | MT Bench⬇️ | IFEval |
---|---|---|
zephyr-7b-gemma-v0.1 | 7.81 | 28.76 |
zephyr-7b-beta | 7.34 | 43.81 |
google/gemma-7b-it | 6.38 | 38.01 |
模型 | AGIEval | GPT4All | TruthfulQA | BigBench | Average⬇️ |
---|---|---|---|---|---|
zephyr-7b-beta | 37.52 | 71.77 | 55.26 | 39.77 | 51.08 |
zephyr-7b-gemma-v0.1 | 34.22 | 66.37 | 52.19 | 37.10 | 47.47 |
mlabonne/Gemmalpaca-7B | 21.6 | 40.87 | 44.85 | 30.49 | 34.45 |
google/gemma-7b-it | 21.33 | 40.84 | 41.70 | 30.25 | 33.53 |
关于 AGIEval、GPT4All、TruthfulQA 和 BigBench 的详细信息
### AGIEval | Task |Version| Metric |Value| |Stderr| |------------------------------|------:|--------|----:|---|-----:| |agieval_aqua_rat | 0|acc |21.65|± | 2.59| | | |acc_norm|25.20|± | 2.73| |agieval_logiqa_en | 0|acc |34.72|± | 1.87| | | |acc_norm|35.94|± | 1.88| |agieval_lsat_ar | 0|acc |19.57|± | 2.62| | | |acc_norm|21.74|± | 2.73| |agieval_lsat_lr | 0|acc |30.59|± | 2.04| | | |acc_norm|32.55|± | 2.08| |agieval_lsat_rc | 0|acc |49.07|± | 3.05| | | |acc_norm|42.75|± | 3.02| |agieval_sat_en | 0|acc |54.85|± | 3.48| | | |acc_norm|53.40|± | 3.48| |agieval_sat_en_without_passage| 0|acc |37.38|± | 3.38| | | |acc_norm|33.98|± | 3.31| |agieval_sat_math | 0|acc |30.91|± | 3.12| | | |acc_norm|28.18|± | 3.04| Average: 34.22% ### GPT4All | Task |Version| Metric |Value| |Stderr| |-------------|------:|------|----:|---|-----:| |arc_challenge| 0|acc |49.15|± | 1.46| | | |acc_norm|52.47|± | 1.46| |arc_easy | 0|acc |77.44|± | 0.86| | | |acc_norm|74.75|± | 0.89| |boolq | 1|acc |79.69|± | 0.70| | | |acc_norm|78.00|± | 0.41| |hellaswag | 0|acc |60.59|± | 0.49| | | |acc_norm|78.00|± | 0.41| |openbookqa | 0|acc |29.20|± | 2.04| | | |acc_norm|37.80|± | 2.17| |piqa | 0|acc |76.82|± | 0.98| | | |acc_norm|77.80|± | 0.97| |winogrande | 0|acc |64.09|± | 1.35| | | |acc_norm|64.09|± | 1.35| Average: 66.37% ### TruthfulQA | Task |Version|Metric|Value| |Stderr| |-------------|------:|------|----:|---|-----:| |truthfulqa_mc| 1|mc1 |35.74|± | 1.68| | | |mc2 |52.19|± | 1.59| Average: 52.19% ### Bigbench | Task |Version| Metric |Value| |Stderr| |------------------------------------------------|------:|---------------------|----:|---|-----:| |bigbench_causal_judgement | 0|multiple_choice_grade|53.68|± | 3.63| |bigbench_date_understanding | 0|multiple_choice_grade|59.89|± | 2.55| |bigbench_disambiguation_qa | 0|multiple_choice_grade|30.23|± | 2.86| |bigbench_geometric_shapes | 0|multiple_choice_grade|11.42|± | 1.68| | | |exact_str_match | 0.00|± | 0.00| |bigbench_logical_deduction_five_objects | 0|multiple_choice_grade|28.40|± | 2.02| |bigbench_logical_deduction_seven_objects | 0|multiple_choice_grade|19.14|± | 1.49| |bigbench_logical_deduction_three_objects | 0|multiple_choice_grade|44.67|± | 2.88| |bigbench_movie_recommendation | 0|multiple_choice_grade|26.80|± | 1.98| |bigbench_navigate | 0|multiple_choice_grade|50.00|± | 1.58| |bigbench_reasoning_about_colored_objects | 0|multiple_choice_grade|52.75|± | 1.12| |bigbench_ruin_names | 0|multiple_choice_grade|33.04|± | 2.22| |bigbench_salient_translation_error_detection | 0|multiple_choice_grade|33.37|± | 1.49| |bigbench_snarks | 0|multiple_choice_grade|48.62|± | 3.73| |bigbench_sports_understanding | 0|multiple_choice_grade|58.11|± | 1.57| |bigbench_temporal_sequences | 0|multiple_choice_grade|37.20|± | 1.53| |bigbench_tracking_shuffled_objects_five_objects | 0|multiple_choice_grade|20.08|± | 1.13| |bigbench_tracking_shuffled_objects_seven_objects| 0|multiple_choice_grade|15.77|± | 0.87| |bigbench_tracking_shuffled_objects_three_objects| 0|multiple_choice_grade|44.67|± | 2.88| Average: 37.1%预期用途和限制
模型最初是基于包含ChatGPT生成的各种合成对话的DEITA 10K数据集进行微调的。
然后,我们进一步将模型与🤗 TRL的 DPOTrainer
在argilla/dpo-mix-7k数据集上进行对齐,该数据集包含7k个由GPT-4排序的提示和模型完成。因此,该模型可用于聊天,您可以查看我们的演示以测试其功能。
以下是您如何使用来自🤗 Transformers的pipeline()
函数运行该模型的步骤:
# pip install transformers>=4.38.2
# pip install accelerate
import torch
from transformers import pipeline
pipe = pipeline(
"text-generation",
model="HuggingFaceH4/zephyr-7b-gemma-v0.1",
device_map="auto",
torch_dtype=torch.bfloat16,
)
messages = [
{
"role": "system",
"content": "", # Model not yet trained for follow this
},
{"role": "user", "content": "How many helicopters can a human eat in one sitting?"},
]
outputs = pipe(
messages,
max_new_tokens=128,
do_sample=True,
temperature=0.7,
top_k=50,
top_p=0.95,
stop_sequence="<|im_end|>",
)
print(outputs[0]["generated_text"][-1]["content"])
# It is not possible for a human to eat a helicopter in one sitting, as a
# helicopter is a large and inedible machine. Helicopters are made of metal,
# plastic, and other materials that are not meant to be consumed by humans.
# Eating a helicopter would be extremely dangerous and would likely cause
# serious health problems, including choking, suffocation, and poisoning. It is
# important to only eat food that is safe and intended for human consumption.
偏差、风险和局限性
Zephyr 7B Gemma还没有在RLHF阶段与人类偏好进行对齐以进行安全性和部署带有类似ChatGPT的循环过滤的响应。因此,模型可能会产生问题输出(尤其是在被提示做此操作时)。此外,训练基础模型(google/gemma-7b
)所使用的语料库的大小和组成尚不清楚,但它很可能包含Web数据和书籍、代码等技术资源的混合。请参见StarCoder2模型卡片以获取此类示例。
训练和评估数据
此模型是 HuggingFaceH4/zephyr-7b-gemma-sft-v0.1 在 argilla/dpo-mix-7k 数据集上微调的版本。
在评估集上实现了以下结果:
- 损失:0.4695
- 奖励/选择:-3.3746
- 奖励/拒绝:-4.9715
- 奖励/准确性:0.7188
- 奖励/边际:1.5970
- Logps/拒绝:-459.4853
- Logps/选择:-429.9115
- Logits/拒绝:86.4684
- Logits/选择:92.8200
训练超参数
以下超参数在训练期间使用
- learning_rate: 5e-07
- train_batch_size: 2
- eval_batch_size: 4
- seed: 42
- distributed_type: multi-GPU
- num_devices: 8
- gradient_accumulation_steps: 8
- total_train_batch_size: 128
- total_eval_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 2
训练结果
训练损失 | 纪元 | 步骤 | 验证损失 | 奖励/选择 | 奖励/拒绝 | 奖励/准确性 | 奖励/边际 | Logps/拒绝 | Logps/选择 | Logits/拒绝 | Logits/选择 |
---|---|---|---|---|---|---|---|---|---|---|---|
0.1923 | 1.9 | 100 | 0.4736 | -3.4575 | -4.9556 | 0.75 | 1.4980 | -459.1662 | -431.5707 | 86.3863 | 92.7360 |
框架版本
- Transformers 4.39.0.dev0
- Pytorch 2.1.2+cu121
- 数据集 2.14.6
- 标记器 0.15.1
引用信息
如果您在工作中发现此模型有用,请考虑引用Zephyr技术报告。
@misc{tunstall2023zephyr,
title={Zephyr: Direct Distillation of LM Alignment},
author={Lewis Tunstall and Edward Beeching and Nathan Lambert and Nazneen Rajani and Kashif Rasul and Younes Belkada and Shengyi Huang and Leandro von Werra and Clémentine Fourrier and Nathan Habib and Nathan Sarrazin and Omar Sanseviero and Alexander M. Rush and Thomas Wolf},
year={2023},
eprint={2310.16944},
archivePrefix={arXiv},
primaryClass={cs.LG}
}
您还可能希望引用此模型的创建者。
@misc{zephyr_7b_gemma,
author = {Lewis Tunstall and Philipp Schmid},
title = {Zephyr 7B Gemma},
year = {2024},
publisher = {Hugging Face},
journal = {Hugging Face repository},
howpublished = {\url{https://hugging-face.cn/HuggingFaceH4/zephyr-7b-gemma-v0.1}}
}
开源LLM排行榜评估结果
详细信息可在此处找到:此处
指标 | 值 |
---|---|
平均值 | 62.41 |
AI2推理挑战(25-Shot) | 58.45 |
HellaSwag(10-Shot) | 83.48 |
MMLU(5-Shot) | 60.68 |
TruthfulQA(0-shot) | 52.07 |
Winogrande(5-shot) | 74.19 |
GSM8k(5-shot) | 45.56 |