sablo-pebble-mistral-dpo-lora-HelpSteer_binarized-2

This model is a fine-tuned version of sablo/sablo-pebble-mistral on the sablo/HelpSteer_binarized dataset. It achieves the following results on the evaluation set:

• Loss: 0.5195
• Rewards/chosen: -1.3821
• Rewards/rejected: -2.4510
• Rewards/accuracies: 0.7358
• Rewards/margins: 1.0689
• Logps/rejected: -158.5470
• Logps/chosen: -147.7195
• Logits/rejected: -2.0952
• Logits/chosen: -2.1922

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-06
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• 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
0.65	0.2	200	0.6563	0.1070	0.0177	0.6509	0.0893	-76.2561	-98.0835	-2.0464	-2.1421
0.456	0.39	400	0.5446	-1.2305	-1.8748	0.7217	0.6444	-139.3410	-142.6661	-2.1203	-2.2102
0.4388	0.59	600	0.5325	-1.8012	-2.8927	0.7123	1.0915	-173.2708	-161.6904	-2.1017	-2.1954
0.6137	0.79	800	0.5198	-1.4487	-2.5199	0.7382	1.0712	-160.8413	-149.9388	-2.0962	-2.1935
0.5866	0.98	1000	0.5195	-1.3821	-2.4510	0.7358	1.0689	-158.5470	-147.7195	-2.0952	-2.1922

Framework versions

• PEFT 0.7.1
• Transformers 4.36.2
• Pytorch 2.0.1+cu118
• Datasets 2.14.6
• Tokenizers 0.15.0