Quantization made by Richard Erkhov.

MiniPLM-Qwen-500M - GGUF

Model creator: https://huggingface.co/MiniLLM/
Original model: https://huggingface.co/MiniLLM/MiniPLM-Qwen-500M/

Name	Quant method	Size
MiniPLM-Qwen-500M.Q2_K.gguf	Q2_K	0.23GB
MiniPLM-Qwen-500M.Q3_K_S.gguf	Q3_K_S	0.25GB
MiniPLM-Qwen-500M.Q3_K.gguf	Q3_K	0.26GB
MiniPLM-Qwen-500M.Q3_K_M.gguf	Q3_K_M	0.26GB
MiniPLM-Qwen-500M.Q3_K_L.gguf	Q3_K_L	0.28GB
MiniPLM-Qwen-500M.IQ4_XS.gguf	IQ4_XS	0.28GB
MiniPLM-Qwen-500M.Q4_0.gguf	Q4_0	0.29GB
MiniPLM-Qwen-500M.IQ4_NL.gguf	IQ4_NL	0.29GB
MiniPLM-Qwen-500M.Q4_K_S.gguf	Q4_K_S	0.29GB
MiniPLM-Qwen-500M.Q4_K.gguf	Q4_K	0.3GB
MiniPLM-Qwen-500M.Q4_K_M.gguf	Q4_K_M	0.3GB
MiniPLM-Qwen-500M.Q4_1.gguf	Q4_1	0.3GB
MiniPLM-Qwen-500M.Q5_0.gguf	Q5_0	0.32GB
MiniPLM-Qwen-500M.Q5_K_S.gguf	Q5_K_S	0.32GB
MiniPLM-Qwen-500M.Q5_K.gguf	Q5_K	0.33GB
MiniPLM-Qwen-500M.Q5_K_M.gguf	Q5_K_M	0.33GB
MiniPLM-Qwen-500M.Q5_1.gguf	Q5_1	0.34GB
MiniPLM-Qwen-500M.Q6_K.gguf	Q6_K	0.36GB
MiniPLM-Qwen-500M.Q8_0.gguf	Q8_0	0.47GB

Original model description:

library_name: transformers license: apache-2.0 datasets: - monology/pile-uncopyrighted - MiniLLM/pile-diff_samp-qwen_1.8B-qwen_104M-r0.5 language: - en metrics: - accuracy pipeline_tag: text-generation

MinPLM-Qwen-500M

paper | code

MiniPLM-Qwen-500M is a 500M model with Qwen achitecture pre-trained from scratch on the Pile using the MiniPLM knowledge distillation framework with the offcial Qwen1.5-1.8B as the teacher model.

We also open-source the pre-training corpus refined by Difference Sampling in MiniPLM for reproducibility.

Evaluation

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

Baseline Models

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}
}