bartowski/Yi-Coder-1.5B-GGUF-torrent

bartowski/Yi-Coder-1.5B-GGUF · Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
FilenameQuant typeFile SizeSplitDescription
Yi-Coder-1.5B-f16.gguff162.95GBfalseFull F16 weights.
Yi-Coder-1.5B-Q8_0.ggufQ8_01.57GBfalseExtremely high quality, generally unneeded but max available quant.
Yi-Coder-1.5B-Q6_K_L.ggufQ6_K_L1.34GBfalseUses Q8_0 for embed and output weights. Very high quality, near perfect, recommended.
Yi-Coder-1.5B-Q6_K.ggufQ6_K1.28GBfalseVery high quality, near perfect, recommended.
Yi-Coder-1.5B-Q5_K_L.ggufQ5_K_L1.18GBfalseUses Q8_0 for embed and output weights. High quality, recommended.
Yi-Coder-1.5B-Q5_K_M.ggufQ5_K_M1.10GBfalseHigh quality, recommended.
Yi-Coder-1.5B-Q4_K_L.ggufQ4_K_L1.06GBfalseUses Q8_0 for embed and output weights. Good quality, recommended.
Yi-Coder-1.5B-Q5_K_S.ggufQ5_K_S1.05GBfalseHigh quality, recommended.
Yi-Coder-1.5B-Q4_K_M.ggufQ4_K_M0.96GBfalseGood quality, default size for must use cases, recommended.
Yi-Coder-1.5B-Q3_K_XL.ggufQ3_K_XL0.94GBfalseUses Q8_0 for embed and output weights. Lower quality but usable, good for low RAM availability.
Yi-Coder-1.5B-Q4_K_S.ggufQ4_K_S0.90GBfalseSlightly lower quality with more space savings, recommended.
Yi-Coder-1.5B-Q4_0_8_8.ggufQ4_0_8_80.87GBfalseOptimized for ARM inference. Requires 'sve' support (see link below).
Yi-Coder-1.5B-Q4_0_4_8.ggufQ4_0_4_80.87GBfalseOptimized for ARM inference. Requires 'i8mm' support (see link below).
Yi-Coder-1.5B-Q4_0_4_4.ggufQ4_0_4_40.87GBfalseOptimized for ARM inference. Should work well on all ARM chips, pick this if you're unsure.
Yi-Coder-1.5B-Q4_0.ggufQ4_00.87GBfalseLegacy format, generally not worth using over similarly sized formats
Yi-Coder-1.5B-IQ4_XS.ggufIQ4_XS0.83GBfalseDecent quality, smaller than Q4_K_S with similar performance, recommended.
Yi-Coder-1.5B-Q3_K_L.ggufQ3_K_L0.83GBfalseLower quality but usable, good for low RAM availability.
Yi-Coder-1.5B-IQ3_M.ggufIQ3_M0.75GBfalseMedium-low quality, new method with decent performance comparable to Q3_K_M.