Requant with iMatrix of : https://huggingface.co/NeverSleepHistorical/MiquMaid-v2-70B-alpha-GGUF
From Q4_K_M through Q8_0.
Q3_K_M quant available, IQ2_XS otw.
For testing purpose, so the folks with 36GB & 24 GB VRAM can use the model.
Some LlamaCPP benchs :
My requant of Miqu :
- Miqu-1-70b-Requant-b1989-iMat-c32_ch400-Q3_K_M.gguf,-,Hellaswag,88.75,,400,2024-01-29 00:00:00,,70b,Mistral_Medium,32768,,,GGUF,- Miqudev,Nexesenex,
- Miqu-1-70b-Requant-b1989-iMat-c32_ch400-Q3_K_M.gguf,-,Hellaswag,88.1,,1000,2024-01-29 00:00:00,,70b,Mistral_Medium,32768,,,GGUF,- Miqudev,Nexesenex,
- Miqu-1-70b-Requant-b1989-iMat-c32_ch400-Q3_K_M.gguf,-,Hellaswag,87.3,,2000,2024-01-29 00:00:00,,70b,Mistral_Medium,32768,,,GGUF,- Miqudev,Nexesenex,
- Miqu-1-70b-Requant-b1989-iMat-c32_ch400-Q3_K_M.gguf,-,Hellaswag_Bin,82,,400,2024-01-29 00:00:00,,70b,Mistral_Medium,32768,,,GGUF,- Miqudev,Nexesenex,
- Miqu-1-70b-Requant-b1989-iMat-c32_ch400-Q3_K_M.gguf,-,Hellaswag_Bin,85.1,,1000,2024-01-29 00:00:00,,70b,Mistral_Medium,32768,,,GGUF,- Miqudev,Nexesenex,
- Miqu-1-70b-Requant-b1989-iMat-c32_ch400-Q3_K_M.gguf,-,Hellaswag_Bin,84.85,,2000,2024-01-29 00:00:00,,70b,Mistral_Medium,32768,,,GGUF,- Miqudev,Nexesenex,
- Miqu-1-70b-Requant-b1989-iMat-c32_ch400-Q3_K_M.gguf,-,Arc-Challenge,57.19063545,,299,2024-01-29 05:40:00,,70b,Mistral_Medium,32768,,,GGUF,- Miqudev,Nexesenex,
- Miqu-1-70b-Requant-b1989-iMat-c32_ch400-Q3_K_M.gguf,-,Arc-Easy,77.19298246,,570,2024-01-29 05:40:00,,70b,Mistral_Medium,32768,,,GGUF,- Miqudev,Nexesenex,
- Miqu-1-70b-Requant-b1989-iMat-c32_ch400-Q3_K_M.gguf,-,MMLU,50.15974441,,313,2024-01-29 05:40:00,,70b,Mistral_Medium,32768,,,GGUF,- Miqudev,Nexesenex,
- Miqu-1-70b-Requant-b1989-iMat-c32_ch400-Q3_K_M.gguf,-,Thruthful-QA,41.49326805,,817,2024-01-29 05:40:00,,70b,Mistral_Medium,32768,,,GGUF,- Miqudev,Nexesenex,
- Miqu-1-70b-Requant-b1989-iMat-c32_ch400-Q3_K_M.gguf,-,Winogrande,78.8477,,1267,2024-01-29 05:40:00,,70b,Mistral_Medium,32768,,,GGUF,- Miqudev,Nexesenex,
- Miqu-1-70b-Requant-b1989-iMat-c32_ch400-Q3_K_M.gguf,-,wikitext,4.2957,512,512,2024-01-29 00:00:00,RBF1000000,70b,Mistral_Medium,32768,,,GGUF,- Miqudev,Nexesenex,81
- Miqu-1-70b-Requant-b1989-iMat-c32_ch400-Q3_K_M.gguf,-,wikitext,3.8380,512,512,2024-01-29 00:00:00,RBF1000000,70b,Mistral_Medium,32768,,,GGUF,- Miqudev,Nexesenex,655
Miqumaid v1 :
- MiquMaid-v1-70B.q3_k_m.gguf,-,Hellaswag,89.25,83.75,400,2024-02-01 00:00:00,,70b,Mistral_Medium,32768,,,GGUF,NeverSleep,NeverSleep,
- MiquMaid-v1-70B.q3_k_m.gguf,-,Hellaswag,88,,1000,2024-02-01 00:00:00,,70b,Mistral_Medium,32768,,,GGUF,NeverSleep,NeverSleep,
- MiquMaid-v1-70B.q3_k_m.gguf,-,Hellaswag_Bin,82.25,,400,2024-02-01 00:00:00,,70b,Mistral_Medium,32768,,,GGUF,NeverSleep,NeverSleep,
- MiquMaid-v1-70B.q3_k_m.gguf,-,Hellaswag_Bin,85,,1000,2024-02-01 00:00:00,,70b,Mistral_Medium,32768,,,GGUF,NeverSleep,NeverSleep,
- MiquMaid-v1-70B.q3_k_m.gguf,-,Arc-Challenge,56.85618729,,299,2024-02-01 05:40:00,,70b,Mistral_Medium,32768,,,GGUF,NeverSleep,NeverSleep,
- MiquMaid-v1-70B.q3_k_m.gguf,-,Arc-Easy,76.31578947,,570,2024-02-01 05:40:00,,70b,Mistral_Medium,32768,,,GGUF,NeverSleep,NeverSleep,
- MiquMaid-v1-70B.q3_k_m.gguf,-,MMLU,49.84025559,,313,2024-02-01 05:40:00,,70b,Mistral_Medium,32768,,,GGUF,NeverSleep,NeverSleep,
- MiquMaid-v1-70B.q3_k_m.gguf,-,Thruthful-QA,40.88127295,,817,2024-02-01 05:40:00,,70b,Mistral_Medium,32768,,,GGUF,NeverSleep,NeverSleep,
- MiquMaid-v1-70B.q3_k_m.gguf,-,Winogrande,79.0845,,1267,2024-02-01 05:40:00,,70b,Mistral_Medium,32768,,,GGUF,NeverSleep,NeverSleep,
- MiquMaid-v1-70B.q3_k_m.gguf,-,wikitext,3.8595,512,512,2024-02-01 00:00:00,70b,Mistral_Medium,32768,,,GGUF,NeverSleep,NeverSleep,655
Miqu DPO :
- Miqu-70B-DPO.q3_k_m.gguf,-,Hellaswag,84.5,400,,2024-02-07 00:00:00,,70b,Mistral_Medium,32768,,,GGUF,NeverSleep,NeverSleep,
- Miqu-70B-DPO.q3_k_m.gguf,-,Hellaswag,83.8,1000,,2024-02-07 00:00:00,,70b,Mistral_Medium,32768,,,GGUF,NeverSleep,NeverSleep,
- Miqu-70B-DPO.q3_k_m.gguf,-,Arc-Challenge,57.85953177,,299,2024-02-07 05:40:00,,70b,Mistral_Medium,32768,,,GGUF,NeverSleep,NeverSleep,
- Miqu-70B-DPO.q3_k_m.gguf,-,Arc-Easy,77.36842105,,570,2024-02-07 05:40:00,,70b,Mistral_Medium,32768,,,GGUF,NeverSleep,NeverSleep,
- Miqu-70B-DPO.q3_k_m.gguf,-,MMLU,50.15974441,,313,2024-02-07 05:40:00,,70b,Mistral_Medium,32768,,,GGUF,NeverSleep,NeverSleep,
- Miqu-70B-DPO.q3_k_m.gguf,-,Thruthful-QA,42.47246022,,817,2024-02-07 05:40:00,,70b,Mistral_Medium,32768,,,GGUF,NeverSleep,NeverSleep,
- Miqu-70B-DPO.q3_k_m.gguf,-,Winogrande,78.7687,,1267,2024-02-07 05:40:00,,70b,Mistral_Medium,32768,,,GGUF,NeverSleep,NeverSleep,
- Miqu-70B-DPO.q3_k_m.gguf,-,wikitext,4.3018,512,512,2024-02-07 00:00:00,,70b,Mistral_Medium,32768,,,GGUF,NeverSleep,NeverSleep,81
- Miqu-70B-DPO.q3_k_m.gguf,-,wikitext,3.8469,512,512,2024-02-07 00:00:00,,70b,Mistral_Medium,32768,,,GGUF,NeverSleep,NeverSleep,655
Miqumaid v2 Alpha Requant :
- MiquMaid-v2-70B-alpha-Requant-b2091-iMat-c32_ch200-Q3_K_M.gguf,-,Hellaswag,84.5,,400,2024-02-07 00:00:00,,70b,Mistral_Medium,32768,,,GGUF,NeverSleep,Nexesenex,
- MiquMaid-v2-70B-alpha-Requant-b2091-iMat-c32_ch200-Q3_K_M.gguf,-,Hellaswag,83.8,,1000,2024-02-07 00:00:00,,70b,Mistral_Medium,32768,,,GGUF,NeverSleep,Nexesenex,
- MiquMaid-v2-70B-alpha-Requant-b2091-iMat-c32_ch200-Q3_K_M.gguf,-,Arc-Challenge,56.18729097,,299,2024-02-07 05:40:00,,70b,Mistral_Medium,32768,,,GGUF,NeverSleep,Nexesenex,
- MiquMaid-v2-70B-alpha-Requant-b2091-iMat-c32_ch200-Q3_K_M.gguf,-,Arc-Easy,77.71929825,,570,2024-02-07 05:40:00,,70b,Mistral_Medium,32768,,,GGUF,NeverSleep,Nexesenex,
- MiquMaid-v2-70B-alpha-Requant-b2091-iMat-c32_ch200-Q3_K_M.gguf,-,MMLU,47.92332268,,313,2024-02-07 05:40:00,,70b,Mistral_Medium,32768,,,GGUF,NeverSleep,Nexesenex,
- MiquMaid-v2-70B-alpha-Requant-b2091-iMat-c32_ch200-Q3_K_M.gguf,-,Thruthful-QA,40.39167687,,817,2024-02-07 05:40:00,,70b,Mistral_Medium,32768,,,GGUF,NeverSleep,Nexesenex,
- MiquMaid-v2-70B-alpha-Requant-b2091-iMat-c32_ch200-Q3_K_M.gguf,-,Winogrande,78.9266,,1267,2024-02-07 05:40:00,,70b,Mistral_Medium,32768,,,GGUF,NeverSleep,Nexesenex,
- MiquMaid-v2-70B-alpha-Requant-b2091-iMat-c32_ch200-Q3_K_M.gguf,-,wikitext,4.4132,512,512,2024-02-07 00:00:00,,70b,Mistral_Medium,32768,,,GGUF,NeverSleep,Nexesenex,81
- MiquMaid-v2-70B-alpha-Requant-b2091-iMat-c32_ch200-Q3_K_M.gguf,-,wikitext,3.9105,512,512,2024-02-07 00:00:00,70b,Mistral_Medium,32768,,,GGUF,NeverSleep,Nexesenex,655
The Hellaswag scores are divergent due to a change in LlamaCPP 5-6 days ago, but in fact, they are stable.
The divergence is showed in Miqumaid v1, which goes from 89.25 to 83.75 on the exact same eval.