|
|
--- |
|
|
library_name: transformers |
|
|
license: llama3.1 |
|
|
base_model: meta-llama/Llama-3.1-8B |
|
|
language: en |
|
|
datasets: |
|
|
- Word2Li/MiddOptimized |
|
|
tags: |
|
|
- llama-factory |
|
|
- full |
|
|
pipeline_tag: text-generation |
|
|
model-index: |
|
|
- name: Llama3.1-8B-Middo-Wizard |
|
|
results: |
|
|
- task: |
|
|
type: text-generation |
|
|
dataset: |
|
|
name: MMLU |
|
|
type: MMLU |
|
|
metrics: |
|
|
- name: weighted accuracy |
|
|
type: weighted accuracy |
|
|
value: 48.39 |
|
|
verified: true |
|
|
- task: |
|
|
type: text-generation |
|
|
dataset: |
|
|
name: IFEval |
|
|
type: IFEval |
|
|
metrics: |
|
|
- name: overall accuracy |
|
|
type: overall accuracy |
|
|
value: 50.11 |
|
|
verified: true |
|
|
- task: |
|
|
type: text-generation |
|
|
dataset: |
|
|
name: GSM8K |
|
|
type: GSM8K |
|
|
metrics: |
|
|
- name: accuracy |
|
|
type: accuracy |
|
|
value: 54.44 |
|
|
verified: true |
|
|
- task: |
|
|
type: text-generation |
|
|
dataset: |
|
|
name: MATH |
|
|
type: MATH |
|
|
metrics: |
|
|
- name: accuracy |
|
|
type: accuracy |
|
|
value: 13.80 |
|
|
verified: true |
|
|
- task: |
|
|
type: text-generation |
|
|
dataset: |
|
|
name: HumanEval |
|
|
type: HumanEval |
|
|
metrics: |
|
|
- name: humaneval_pass@1 |
|
|
type: humaneval_pass@1 |
|
|
value: 46.95 |
|
|
verified: true |
|
|
- task: |
|
|
type: text-generation |
|
|
dataset: |
|
|
name: MBPP |
|
|
type: MBPP |
|
|
metrics: |
|
|
- name: score |
|
|
type: score |
|
|
value: 45.00 |
|
|
verified: true |
|
|
- task: |
|
|
type: text-generation |
|
|
dataset: |
|
|
name: Hellaswag |
|
|
type: Hellaswag |
|
|
metrics: |
|
|
- name: accuracy |
|
|
type: accuracy |
|
|
value: 63.54 |
|
|
verified: true |
|
|
- task: |
|
|
type: text-generation |
|
|
dataset: |
|
|
name: GPQA |
|
|
type: GPQA |
|
|
metrics: |
|
|
- name: accuracy |
|
|
type: accuracy |
|
|
value: 20.20 |
|
|
verified: true |
|
|
metrics: |
|
|
- accuracy |
|
|
--- |
|
|
|
|
|
# Llama3.1-8B-Middo-Wizard |
|
|
|
|
|
Paper: [Middo: Model-Informed Dynamic Data Optimization for Enhanced LLM Fine-Tuning via Closed-Loop Learning](https://arxiv.org/abs/2508.21589) |
|
|
|
|
|
Code: https://github.com/Word2VecT/Middo |
|
|
|
|
|
## Model description |
|
|
|
|
|
This model is a fine-tuned version of [meta-llama/Llama-3.1-8B](https://huggingface.co/meta-llama/Llama-3.1-8B) on the [MiddOptimzed/llama_wizard](https://huggingface.co/datasets/Word2Li/MiddOptimized/viewer/default/llama_alpaca) dataset. |
|
|
|
|
|
## Training and evaluation data |
|
|
|
|
|
### Training data |
|
|
|
|
|
Middo optimized [WizardLM_evol_instruct_70k](https://huggingface.co/datasets/WizardLMTeam/WizardLM_evol_instruct_70k) on [meta-llama/Llama-3.1-8B](https://huggingface.co/meta-llama/Llama-3.1-8B). |
|
|
|
|
|
### Evaluation data |
|
|
|
|
|
- General |
|
|
- MMLU |
|
|
- IFEval |
|
|
- Math |
|
|
- GSM8K |
|
|
- MATH |
|
|
- Code |
|
|
- HumanEval |
|
|
- MBPP |
|
|
- Reasoning |
|
|
- Hellaswag |
|
|
- GPQA |
|
|
|
|
|
## Training procedure |
|
|
|
|
|
### Training hyperparameters |
|
|
|
|
|
The following hyperparameters were used during training: |
|
|
|
|
|
- learning_rate: 2e-05 |
|
|
- train_batch_size: 4 |
|
|
- eval_batch_size: 8 |
|
|
- seed: 42 |
|
|
- distributed_type: multi-GPU |
|
|
- num_devices: 8 |
|
|
- gradient_accumulation_steps: 8 |
|
|
- total_train_batch_size: 256 |
|
|
- total_eval_batch_size: 64 |
|
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
|
- lr_scheduler_type: cosine |
|
|
- lr_scheduler_warmup_ratio: 0.03 |
|
|
- num_epochs: 1.0 |
|
|
|
|
|
### Training results |
|
|
|
|
|
- epoch: 0.9973935708079931 |
|
|
- total_flos: 2.698045158024282e + 18 |
|
|
- train_loss: 0.5919382667707649 |
|
|
- train_runtime: 4471.5794 |
|
|
- train_samples_per_second: 16.469 |
|
|
- train_steps_per_second: 0.064 |
|
|
|
|
|
### Framework versions |
|
|
|
|
|
- Transformers 4.45.2 |
|
|
- Pytorch 2.5.1+cu121 |
|
|
- Datasets 2.21.0 |
|
|
- Tokenizers 0.20.1 |
|
|
|
