RichardErkhov/IntelLabs_-_sqft-phi-3-mini-4k-50-base-gguf

Quantization made by Richard Erkhov.

Github

Discord

Request more models

sqft-phi-3-mini-4k-50-base - GGUF

• Model creator: https://huggingface.co/IntelLabs/
• Original model: https://huggingface.co/IntelLabs/sqft-phi-3-mini-4k-50-base/

Name	Quant method	Size
sqft-phi-3-mini-4k-50-base.Q2_K.gguf	Q2_K	1.32GB
sqft-phi-3-mini-4k-50-base.IQ3_XS.gguf	IQ3_XS	1.51GB
sqft-phi-3-mini-4k-50-base.IQ3_S.gguf	IQ3_S	1.57GB
sqft-phi-3-mini-4k-50-base.Q3_K_S.gguf	Q3_K_S	1.57GB
sqft-phi-3-mini-4k-50-base.IQ3_M.gguf	IQ3_M	1.73GB
sqft-phi-3-mini-4k-50-base.Q3_K.gguf	Q3_K	1.82GB
sqft-phi-3-mini-4k-50-base.Q3_K_M.gguf	Q3_K_M	1.82GB
sqft-phi-3-mini-4k-50-base.Q3_K_L.gguf	Q3_K_L	1.94GB
sqft-phi-3-mini-4k-50-base.IQ4_XS.gguf	IQ4_XS	1.93GB
sqft-phi-3-mini-4k-50-base.Q4_0.gguf	Q4_0	2.03GB
sqft-phi-3-mini-4k-50-base.IQ4_NL.gguf	IQ4_NL	2.04GB
sqft-phi-3-mini-4k-50-base.Q4_K_S.gguf	Q4_K_S	2.04GB
sqft-phi-3-mini-4k-50-base.Q4_K.gguf	Q4_K	2.23GB
sqft-phi-3-mini-4k-50-base.Q4_K_M.gguf	Q4_K_M	2.23GB
sqft-phi-3-mini-4k-50-base.Q4_1.gguf	Q4_1	2.24GB
sqft-phi-3-mini-4k-50-base.Q5_0.gguf	Q5_0	2.46GB
sqft-phi-3-mini-4k-50-base.Q5_K_S.gguf	Q5_K_S	2.46GB
sqft-phi-3-mini-4k-50-base.Q5_K.gguf	Q5_K	2.62GB
sqft-phi-3-mini-4k-50-base.Q5_K_M.gguf	Q5_K_M	2.62GB
sqft-phi-3-mini-4k-50-base.Q5_1.gguf	Q5_1	2.68GB
sqft-phi-3-mini-4k-50-base.Q6_K.gguf	Q6_K	2.92GB
sqft-phi-3-mini-4k-50-base.Q8_0.gguf	Q8_0	3.78GB

Original model description:

language: en license: apache-2.0

SQFT Base Model: sqft-phi-3-mini-4k-50-base

• Source Model: microsoft/Phi-3-mini-4k-instruct
• Sparse Method: Wanda
• Sparsity: 50%
• Quantization: No

Model Sources

• Repository: https://github.com/IntelLabs/Hardware-Aware-Automated-Machine-Learning/tree/main/SQFT
• Paper: SQFT: Low-cost Model Adaptation in Low-precision Sparse Foundation Models

How to get this model

Refer to the command in SQFT/run_command/phi-3-mini-4k-instruct/sparse_quantization.sh#11.

Citation

@article{munoz2024sqft,
  title = {SQFT: Low-cost Model Adaptation in Low-precision Sparse Foundation Models},
  author={J. Pablo Munoz and Jinjie Yuan and Nilesh Jain},
  journal={},
  year={2024}
}

Acknowledgement

Thanks to the work Wanda (paper, code), which provides a simple but effective pruning approach.

License

Apache-2.0