Quantization made by Richard Erkhov.

Ref-Pretrain-Qwen-104M - GGUF

Model creator: https://huggingface.co/MiniLLM/
Original model: https://huggingface.co/MiniLLM/Ref-Pretrain-Qwen-104M/

Name	Quant method	Size
Ref-Pretrain-Qwen-104M.Q2_K.gguf	Q2_K	0.07GB
Ref-Pretrain-Qwen-104M.IQ3_XS.gguf	IQ3_XS	0.08GB
Ref-Pretrain-Qwen-104M.IQ3_S.gguf	IQ3_S	0.08GB
Ref-Pretrain-Qwen-104M.Q3_K_S.gguf	Q3_K_S	0.08GB
Ref-Pretrain-Qwen-104M.IQ3_M.gguf	IQ3_M	0.08GB
Ref-Pretrain-Qwen-104M.Q3_K.gguf	Q3_K	0.08GB
Ref-Pretrain-Qwen-104M.Q3_K_M.gguf	Q3_K_M	0.08GB
Ref-Pretrain-Qwen-104M.Q3_K_L.gguf	Q3_K_L	0.08GB
Ref-Pretrain-Qwen-104M.IQ4_XS.gguf	IQ4_XS	0.08GB
Ref-Pretrain-Qwen-104M.Q4_0.gguf	Q4_0	0.08GB
Ref-Pretrain-Qwen-104M.IQ4_NL.gguf	IQ4_NL	0.08GB
Ref-Pretrain-Qwen-104M.Q4_K_S.gguf	Q4_K_S	0.08GB
Ref-Pretrain-Qwen-104M.Q4_K.gguf	Q4_K	0.08GB
Ref-Pretrain-Qwen-104M.Q4_K_M.gguf	Q4_K_M	0.08GB
Ref-Pretrain-Qwen-104M.Q4_1.gguf	Q4_1	0.08GB
Ref-Pretrain-Qwen-104M.Q5_0.gguf	Q5_0	0.08GB
Ref-Pretrain-Qwen-104M.Q5_K_S.gguf	Q5_K_S	0.08GB
Ref-Pretrain-Qwen-104M.Q5_K.gguf	Q5_K	0.08GB
Ref-Pretrain-Qwen-104M.Q5_K_M.gguf	Q5_K_M	0.08GB
Ref-Pretrain-Qwen-104M.Q5_1.gguf	Q5_1	0.08GB
Ref-Pretrain-Qwen-104M.Q6_K.gguf	Q6_K	0.09GB
Ref-Pretrain-Qwen-104M.Q8_0.gguf	Q8_0	0.11GB

Original model description:

library_name: transformers license: apache-2.0 datasets: - monology/pile-uncopyrighted - MiniLLM/pile-tokenized language: - en metrics: - accuracy pipeline_tag: text-generation

Ref-Pretrain-Qwen-104M

paper | code

Ref-Pretrain-Qwen-104M is a 104M model with Qwen achitecture conventionally pre-trained from scratch on the Pile for 5B tokens.

We also open-source the tokenized pre-training corpus for reproducibility.

It is used as the reference model in the MiniPLM knwoledge distillation framework to construct the refined pre-training corpus. The data is then used to train MiniPLM models.

Evaluation

MiniPLM models achieves better performance given the same computation and scales well across model sizes:

Citation

@article{miniplm,
    title={MiniPLM: Knowledge Distillation for Pre-Training Language Models}, 
    author={Yuxian Gu and Hao Zhou and Fandong Meng and Jie Zhou and Minlie Huang},
    journal={arXiv preprint arXiv:2410.17215},
    year={2024}
}