CoolFace
Modelpublic

RichardErkhov/dawn17_-_MaidStarling-2x7B-base-gguf

sourceHugging Faceupdated 2y agoView on Hugging Face
0likes2.7kdownloads
Model Card

Quantization made by Richard Erkhov.

Github

Discord

Request more models

MaidStarling-2x7B-base - GGUF

  • —Model creator: https://huggingface.co/dawn17/
  • —Original model: https://huggingface.co/dawn17/MaidStarling-2x7B-base/
NameQuant methodSize
MaidStarling-2x7B-base.Q2_K.ggufQ2_K4.43GB
MaidStarling-2x7B-base.IQ3_XS.ggufIQ3_XS4.95GB
MaidStarling-2x7B-base.IQ3_S.ggufIQ3_S5.22GB
MaidStarling-2x7B-base.Q3_K_S.ggufQ3KS5.2GB
MaidStarling-2x7B-base.IQ3_M.ggufIQ3_M5.35GB
MaidStarling-2x7B-base.Q3_K.ggufQ3_K5.78GB
MaidStarling-2x7B-base.Q3_K_M.ggufQ3KM5.78GB
MaidStarling-2x7B-base.Q3_K_L.ggufQ3KL6.27GB
MaidStarling-2x7B-base.IQ4_XS.ggufIQ4_XS6.5GB
MaidStarling-2x7B-base.Q4_0.ggufQ4_06.78GB
MaidStarling-2x7B-base.IQ4_NL.ggufIQ4_NL6.85GB
MaidStarling-2x7B-base.Q4_K_S.ggufQ4KS6.84GB
MaidStarling-2x7B-base.Q4_K.ggufQ4_K7.25GB
MaidStarling-2x7B-base.Q4_K_M.ggufQ4KM7.25GB
MaidStarling-2x7B-base.Q4_1.ggufQ4_17.52GB
MaidStarling-2x7B-base.Q5_0.ggufQ5_08.26GB
MaidStarling-2x7B-base.Q5_K_S.ggufQ5KS8.26GB
MaidStarling-2x7B-base.Q5_K.ggufQ5_K8.51GB
MaidStarling-2x7B-base.Q5_K_M.ggufQ5KM8.51GB
MaidStarling-2x7B-base.Q5_1.ggufQ5_19.01GB
MaidStarling-2x7B-base.Q6_K.ggufQ6_K9.84GB
MaidStarling-2x7B-base.Q8_0.ggufQ8_012.75GB

Original model description: --- license: apache-2.0 ---

basemodel: /Users/dawn/git/models/Silicon-Maid-7B gatemode: hidden # one of "hidden", "cheap_embed", or "random" dtype: bfloat16 # output dtype (float32, float16, or bfloat16) experts:

  • —sourcemodel: /Users/dawn/git/models/Silicon-Maid-7B positiveprompts:
  • —"roleplay"
  • —sourcemodel: /Users/dawn/git/models/Starling-LM-7B-beta positiveprompts:
  • —"chat"

Open LLM Leaderboard Evaluation Results

MetricValue
Avg.70.76
AI2 Reasoning Challenge (25-Shot)68.43
HellaSwag (10-Shot)86.28
MMLU (5-Shot)60.34
TruthfulQA (0-shot)60.34
Winogrande (5-shot)78.93
GSM8k (5-shot)65.43