README.md

metadata

license: apache-2.0
library_name: peft
tags:
  - trl
  - orpo
  - generated_from_trainer
base_model: sail/Sailor-1.8B
model-index:
  - name: OrpoSailor-1.8B
    results: []

OrpoSailor-1.8B

This model is a fine-tuned version of sail/Sailor-1.8B on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 4.4472
• Rewards/chosen: -0.1546
• Rewards/rejected: -0.1440
• Rewards/accuracies: 0.0
• Rewards/margins: -0.0106
• Logps/rejected: -1.4399
• Logps/chosen: -1.5456
• Logits/rejected: 0.0181
• Logits/chosen: -1.6644
• Nll Loss: 4.3708
• Log Odds Ratio: -0.7637
• Log Odds Chosen: -0.1364

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: 8e-06
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 10
• 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	Nll Loss	Log Odds Ratio	Log Odds Chosen
4.5248	0.8	1	4.4472	-0.1546	-0.1440	0.0	-0.0106	-1.4399	-1.5456	0.0181	-1.6644	4.3708	-0.7637	-0.1364

Framework versions

• PEFT 0.10.0
• Transformers 4.40.0
• Pytorch 2.1.2
• Datasets 2.19.0
• Tokenizers 0.19.1