spin-v-deduplicate

This model is a fine-tuned version of alignment-handbook/zephyr-7b-sft-full on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0058
• Rewards/real: -1.4306
• Rewards/generated: -15.5402
• Rewards/accuracies: 1.0
• Rewards/margins: 14.1096
• Logps/generated: -246.4990
• Logps/real: -144.4687
• Logits/generated: -2.5886
• Logits/real: -2.1940

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: 4
• total_train_batch_size: 32
• total_eval_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/real	Rewards/generated	Rewards/accuracies	Rewards/margins	Logps/generated	Logps/real	Logits/generated	Logits/real
0.0136	0.32	100	0.0116	0.0287	-9.1564	1.0	9.1852	-182.6614	-129.8758	-2.7849	-2.4559
0.0049	0.64	200	0.0068	-1.1503	-13.1741	1.0	12.0237	-222.8379	-141.6664	-2.6896	-2.3087
0.0026	0.96	300	0.0058	-1.4306	-15.5402	1.0	14.1096	-246.4990	-144.4687	-2.5886	-2.1940

Framework versions

• Transformers 4.37.0
• Pytorch 2.1.2+cu121
• Datasets 2.14.6
• Tokenizers 0.15.2