metadata

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-dpo-full
    results: []

zephyr-7b-dpo-full

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: 0.2141
Rewards/chosen: 4.2427
Rewards/rejected: -5.5362
Rewards/accuracies: 0.9141
Rewards/margins: 9.7789
Logps/rejected: -191.0266
Logps/chosen: -248.9478
Logits/rejected: -2.4882
Logits/chosen: -2.5045

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: 2
total_train_batch_size: 128
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: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.3315	0.21	100	0.2975	3.4373	-3.8279	0.9023	7.2653	-187.6101	-250.5587	-2.5158	-2.5301
0.2909	0.42	200	0.2754	4.8618	-4.1200	0.9180	8.9818	-188.1942	-247.7097	-2.5310	-2.5459
0.6445	0.63	300	0.2245	4.2165	-5.3713	0.9102	9.5878	-190.6968	-249.0004	-2.4915	-2.5059
0.2653	0.84	400	0.2103	4.5125	-5.2808	0.9258	9.7933	-190.5157	-248.4084	-2.4965	-2.5123

Framework versions

Transformers 4.38.2
Pytorch 2.1.2+cu118
Datasets 2.16.1
Tokenizers 0.15.2