zephyr-7b-gemma-dpo/README.md

library_name, tags, model-index

library_name	tags	model-index
transformers	trl dpo generated_from_trainer	name results zephyr-7b-gemma-dpo

zephyr-7b-gemma-dpo

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

• Loss: 0.5287
• Rewards/chosen: 0.4753
• Rewards/rejected: -0.1181
• Rewards/accuracies: 0.7708
• Rewards/margins: 0.5934
• Logps/rejected: -326.4217
• Logps/chosen: -277.1591
• Logits/rejected: -1.3248
• Logits/chosen: -1.3362

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: 4
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 4
• 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: 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
0.5532	0.9858	52	0.5591	0.4980	0.0692	0.7604	0.4288	-322.6760	-276.7043	-1.3205	-1.3316
0.4562	1.9716	104	0.5287	0.4753	-0.1181	0.7708	0.5934	-326.4217	-277.1591	-1.3248	-1.3362

Framework versions

• Transformers 4.45.2
• Pytorch 2.5.1+rocm6.2
• Datasets 3.2.0
• Tokenizers 0.20.3