Models Trained with Random Mixture

This is a collection of 64 language models, each with approximately 1B parameters, trained on different random mixtures of data. This project aims to validate the generalization capabilities of the RegMix approach (https://huggingface.co/papers/2407.01492) from small-scale (e.g., 1M parameters) to large-scale (e.g., 1B parameters) models.

Key Features

• Model Size: 64 separate models, each with ~1B parameters
• Training Data: Random data mixtures on the RegMix-Data dataset
• Purpose: To validate the effectiveness of RegMix on identifying high-performing data mixture

Dataset

The models were trained using the RegMix-Data dataset, which is split into different domains from The Pile dataset.

Training Hyperparameters

How to Load a Model

You can load any model using the corresponding branch with the Hugging Face Transformers library:

from transformers import AutoModel, AutoTokenizer

model = AutoModel.from_pretrained("sail/data-mixture-random-1b", revision="model-index-1")
tokenizer = AutoTokenizer.from_pretrained("sail/data-mixture-random-1b", revision="model-index-1")

Data Mixture

The specific data mixture used for training each 1B model can be found in the file train_config.yaml in each corresponding model branch.

Model Variants

To access different model variants, simply change the revision parameter in the from_pretrained method to the desired model index (e.g., "model-index-2", "model-index-3"), and the maxium index is 64.

Usage Notes

• These models are primarily intended for research purposes.
• Performance may vary depending on the specific task and domain.

Citation

If you use these models in your research, please cite the RegMix paper:

@article{liu2024regmix,
  title={RegMix: Data Mixture as Regression for Language Model Pre-training},
  author={Liu, Qian and Zheng, Xiaosen and Muennighoff, Niklas and Zeng, Guangtao and Dou, Longxu and Pang, Tianyu and Jiang, Jing and Lin, Min},
  journal={arXiv preprint arXiv:2407.01492},
  year={2024}
}

For more information about the RegMix methodology and its applications, please refer to the original paper.

Performance

We evaluated each model using lm-evaluation-harness. The performance metric for each task is the average of 0-shot to 5-shot accnorm (accuracy normalized, if available) or acc (accuracy) scores.

Table 1: Model Index 1-8

Table 2: Model Index 9-16

Table 3: Model Index 17-24

Table 4: Model Index 25-32

Table 5: Model Index 33-40

Table 6: Model Index 41-48

Table 7: Model Index 49-56

Table 8: Model Index 57-64