CoolFace
Modelpublic

RichardErkhov/taicheng_-_zephyr-7b-align-scan-2e-07-0.39-polynomial-1.0-gguf

sourceHugging Faceupdated 1y agoView on Hugging Face
0likes399downloads
Model Card

Quantization made by Richard Erkhov.

Github

Discord

Request more models

zephyr-7b-align-scan-2e-07-0.39-polynomial-1.0 - GGUF

  • —Model creator: https://huggingface.co/taicheng/
  • —Original model: https://huggingface.co/taicheng/zephyr-7b-align-scan-2e-07-0.39-polynomial-1.0/
NameQuant methodSize
zephyr-7b-align-scan-2e-07-0.39-polynomial-1.0.Q2_K.ggufQ2_K2.53GB
zephyr-7b-align-scan-2e-07-0.39-polynomial-1.0.IQ3_XS.ggufIQ3_XS2.81GB
zephyr-7b-align-scan-2e-07-0.39-polynomial-1.0.IQ3_S.ggufIQ3_S2.96GB
zephyr-7b-align-scan-2e-07-0.39-polynomial-1.0.Q3_K_S.ggufQ3KS2.95GB
zephyr-7b-align-scan-2e-07-0.39-polynomial-1.0.IQ3_M.ggufIQ3_M3.06GB
zephyr-7b-align-scan-2e-07-0.39-polynomial-1.0.Q3_K.ggufQ3_K3.28GB
zephyr-7b-align-scan-2e-07-0.39-polynomial-1.0.Q3_K_M.ggufQ3KM3.28GB
zephyr-7b-align-scan-2e-07-0.39-polynomial-1.0.Q3_K_L.ggufQ3KL3.56GB
zephyr-7b-align-scan-2e-07-0.39-polynomial-1.0.IQ4_XS.ggufIQ4_XS3.67GB
zephyr-7b-align-scan-2e-07-0.39-polynomial-1.0.Q4_0.ggufQ4_03.83GB
zephyr-7b-align-scan-2e-07-0.39-polynomial-1.0.IQ4_NL.ggufIQ4_NL3.87GB
zephyr-7b-align-scan-2e-07-0.39-polynomial-1.0.Q4_K_S.ggufQ4KS3.86GB
zephyr-7b-align-scan-2e-07-0.39-polynomial-1.0.Q4_K.ggufQ4_K4.07GB
zephyr-7b-align-scan-2e-07-0.39-polynomial-1.0.Q4_K_M.ggufQ4KM4.07GB
zephyr-7b-align-scan-2e-07-0.39-polynomial-1.0.Q4_1.ggufQ4_14.24GB
zephyr-7b-align-scan-2e-07-0.39-polynomial-1.0.Q5_0.ggufQ5_04.65GB
zephyr-7b-align-scan-2e-07-0.39-polynomial-1.0.Q5_K_S.ggufQ5KS4.65GB
zephyr-7b-align-scan-2e-07-0.39-polynomial-1.0.Q5_K.ggufQ5_K4.78GB
zephyr-7b-align-scan-2e-07-0.39-polynomial-1.0.Q5_K_M.ggufQ5KM4.78GB
zephyr-7b-align-scan-2e-07-0.39-polynomial-1.0.Q5_1.ggufQ5_15.07GB
zephyr-7b-align-scan-2e-07-0.39-polynomial-1.0.Q6_K.ggufQ6_K5.53GB
zephyr-7b-align-scan-2e-07-0.39-polynomial-1.0.Q8_0.ggufQ8_07.17GB

Original model description: --- libraryname: transformers license: apache-2.0 basemodel: alignment-handbook/zephyr-7b-sft-full tags:

  • —alignment-handbook
  • —trl
  • —dpo
  • —generatedfromtrainer
  • —trl
  • —dpo
  • —generatedfromtrainer datasets:
  • —HuggingFaceH4/ultrafeedback_binarized model-index:
  • —name: zephyr-7b-align-scan-2e-07-0.39-polynomial-1.0 results: [] ---

<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->

zephyr-7b-align-scan-2e-07-0.39-polynomial-1.0

This model is a fine-tuned version of alignment-handbook/zephyr-7b-sft-full on the HuggingFaceH4/ultrafeedback_binarized dataset. It achieves the following results on the evaluation set:

  • —Loss: 0.6326
  • —Rewards/chosen: 0.3567
  • —Rewards/rejected: -0.0406
  • —Rewards/accuracies: 0.3492
  • —Rewards/margins: 0.3973
  • —Logps/rejected: -81.2324
  • —Logps/chosen: -73.5766
  • —Logits/rejected: -2.5151
  • —Logits/chosen: -2.5314

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • —learning_rate: 2e-07
  • —trainbatchsize: 8
  • —evalbatchsize: 8
  • —seed: 42
  • —distributed_type: multi-GPU
  • —num_devices: 4
  • —gradientaccumulationsteps: 2
  • —totaltrainbatch_size: 64
  • —totalevalbatch_size: 32
  • —optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • —lrschedulertype: polynomial
  • —lrschedulerwarmup_ratio: 0.1
  • —num_epochs: 1

Training results

Training LossEpochStepValidation LossRewards/chosenRewards/rejectedRewards/accuraciesRewards/marginsLogps/rejectedLogps/chosenLogits/rejectedLogits/chosen
0.65380.34841000.63200.58100.29640.34720.2846-80.3683-73.0014-2.5308-2.5469
0.64620.69692000.63240.2521-0.11280.34920.3648-81.4175-73.8449-2.5055-2.5222

Framework versions

  • —Transformers 4.44.2
  • —Pytorch 2.4.0
  • —Datasets 2.21.0
  • —Tokenizers 0.19.1