Quantization made by Richard Erkhov.

MiniPLM-llama3.1-212M - GGUF

Model creator: https://huggingface.co/MiniLLM/
Original model: https://huggingface.co/MiniLLM/MiniPLM-llama3.1-212M/

Name	Quant method	Size
MiniPLM-llama3.1-212M.Q2_K.gguf	Q2_K	0.12GB
MiniPLM-llama3.1-212M.IQ3_XS.gguf	IQ3_XS	0.13GB
MiniPLM-llama3.1-212M.IQ3_S.gguf	IQ3_S	0.13GB
MiniPLM-llama3.1-212M.Q3_K_S.gguf	Q3_K_S	0.13GB
MiniPLM-llama3.1-212M.IQ3_M.gguf	IQ3_M	0.13GB
MiniPLM-llama3.1-212M.Q3_K.gguf	Q3_K	0.13GB
MiniPLM-llama3.1-212M.Q3_K_M.gguf	Q3_K_M	0.13GB
MiniPLM-llama3.1-212M.Q3_K_L.gguf	Q3_K_L	0.14GB
MiniPLM-llama3.1-212M.IQ4_XS.gguf	IQ4_XS	0.14GB
MiniPLM-llama3.1-212M.Q4_0.gguf	Q4_0	0.14GB
MiniPLM-llama3.1-212M.IQ4_NL.gguf	IQ4_NL	0.14GB
MiniPLM-llama3.1-212M.Q4_K_S.gguf	Q4_K_S	0.14GB
MiniPLM-llama3.1-212M.Q4_K.gguf	Q4_K	0.15GB
MiniPLM-llama3.1-212M.Q4_K_M.gguf	Q4_K_M	0.15GB
MiniPLM-llama3.1-212M.Q4_1.gguf	Q4_1	0.15GB
MiniPLM-llama3.1-212M.Q5_0.gguf	Q5_0	0.16GB
MiniPLM-llama3.1-212M.Q5_K_S.gguf	Q5_K_S	0.16GB
MiniPLM-llama3.1-212M.Q5_K.gguf	Q5_K	0.16GB
MiniPLM-llama3.1-212M.Q5_K_M.gguf	Q5_K_M	0.16GB
MiniPLM-llama3.1-212M.Q5_1.gguf	Q5_1	0.16GB
MiniPLM-llama3.1-212M.Q6_K.gguf	Q6_K	0.17GB
MiniPLM-llama3.1-212M.Q8_0.gguf	Q8_0	0.22GB

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

MiniPLM-llama3.1-212M

paper | code

MiniPLM-llama3.1-212M is a 212M model with the LLaMA3.1 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. This model shows the flexibility of the MiniPLM framework in conducting knowledge distillation across model families.

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

Conventional Pre-Training

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}
}