metadata

library_name: transformers
license: apache-2.0
base_model: alignment-handbook/zephyr-7b-sft-full
tags:
  - alignment-handbook
  - trl
  - dpo
  - generated_from_trainer
  - trl
  - dpo
  - generated_from_trainer
datasets:
  - HuggingFaceH4/ultrafeedback_binarized
model-index:
  - name: zephyr-7b-uf-dpo-2e
    results: []

zephyr-7b-uf-dpo-2e

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: 0.5016
Rewards/chosen: -1.5911
Rewards/rejected: -2.8172
Rewards/accuracies: 0.7695
Rewards/margins: 1.2260
Logps/rejected: -544.3789
Logps/chosen: -421.7447
Logits/rejected: 2.2404
Logits/chosen: 1.4759

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-07
train_batch_size: 8
eval_batch_size: 8
seed: 42
distributed_type: multi-GPU
num_devices: 8
gradient_accumulation_steps: 4
total_train_batch_size: 256
total_eval_batch_size: 64
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.1
num_epochs: 2

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.5692	0.4184	100	0.5671	-0.3691	-0.9295	0.7344	0.5604	-355.6129	-299.5389	-1.1363	-1.3128
0.5056	0.8368	200	0.5104	-0.9550	-1.8507	0.7852	0.8957	-447.7286	-358.1263	1.4954	0.9604
0.3744	1.2552	300	0.5120	-1.3455	-2.5618	0.7617	1.2163	-518.8406	-397.1791	1.8527	0.9871
0.351	1.6736	400	0.5019	-1.5195	-2.7388	0.7695	1.2193	-536.5389	-414.5813	2.1782	1.4005

Framework versions

Transformers 4.44.1
Pytorch 2.1.2+cu121
Datasets 2.21.0
Tokenizers 0.19.1