Teachers_Exam_LLaMA_8B

<!-- 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. -->

sft

This model is a fine-tuned version of /home/Meta-Llama-3-8B-Instruct on the alpaca_zh_demo, the identity and the teachers_exam_local datasets. It achieves the following results on the evaluation set:

• Loss: 1.5271

这个模型是Meta-Llama-3-8B-Instruct在identity和teachers_exam_local数据集上的微调版本。 在评估集上得到如下结果:

• Loss: 1.5271

Model description

This Model is based Meta-Llama3-8B-Instruct finetuning.

Intended uses & limitations

No limit,everyone can use it. 无限制使用，任何人都可以使用。

Training and evaluation data

Used high qulity datas with Teachers Exam.The data include choice,multi-choice,etc types subjects. 使用高质量的数据与教师考试。数据包括选择题、多项选择题等类型的题目。

#大模型评测 benchmark(mmlu)

        Average: 66.89                                                                
           STEM: 57.52
Social Sciences: 76.28
     Humanities: 62.32
          Other: 73.32

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 2
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 100
• num_epochs: 8.0
• mixed_precision_training: Native AMP

Training results

Framework versions

• PEFT 0.12.0
• Transformers 4.43.4
• Pytorch 2.4.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1