RichardErkhov/CharlesLi_-_OpenELM-1_1B-DPO-full-max-second-reward-gguf

Quantization made by Richard Erkhov.

Github

Discord

Request more models

OpenELM-1_1B-DPO-full-max-second-reward - GGUF

• —Model creator: https://huggingface.co/CharlesLi/
• —Original model: https://huggingface.co/CharlesLi/OpenELM-1_1B-DPO-full-max-second-reward/

Original model description: --- library_name: transformers tags:

• —trl
• —dpo
• —alignment-handbook
• —generatedfromtrainer model-index:
• —name: OpenELM-1_1B-DPO-full-max-second-reward 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. -->

OpenELM-1_1B-DPO-full-max-second-reward

This model was trained from scratch on an unknown dataset. It achieves the following results on the evaluation set:

• —Loss: 1.4829
• —Rewards/chosen: -12.4375
• —Rewards/rejected: -12.875
• —Rewards/accuracies: 0.5371
• —Rewards/margins: 0.4414
• —Logps/rejected: -1576.0
• —Logps/chosen: -1560.0
• —Logits/rejected: 10.8125
• —Logits/chosen: 8.8125

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: 5e-05
• —trainbatchsize: 8
• —evalbatchsize: 16
• —seed: 42
• —distributed_type: multi-GPU
• —num_devices: 4
• —gradientaccumulationsteps: 2
• —totaltrainbatch_size: 64
• —totalevalbatch_size: 64
• —optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• —lrschedulertype: cosine
• —lrschedulerwarmup_ratio: 0.1
• —num_epochs: 3

Training results

Framework versions

• —Transformers 4.44.2
• —Pytorch 2.3.0
• —Datasets 3.0.0
• —Tokenizers 0.19.1