DeepSeek-V3 Pruning and Quantization
Collection
11 items
•
Updated
Original model: tflsxyy/DeepSeek-V3-0324-MoE-Pruner-E160-bf16
All quants made using ggerganov/llama.cpp and unslothai/llama.cpp
All quants made using imatrix option with dataset from bartowski1182/calibration_datav3.txt, tristandruyen/calibration_data_v5_rc.txt and bartowski1182/qwen_calibration_with_chat.txt.
4-bit