Na_L3_250steps_1e7rate_03beta_cSFTDPO

This model is a fine-tuned version of tsavage68/Na_L3_100steps_1e6rate_SFT on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.0003
• Rewards/chosen: 1.1615
• Rewards/rejected: -7.2781
• Rewards/accuracies: 1.0
• Rewards/margins: 8.4396
• Logps/rejected: -65.7620
• Logps/chosen: -21.0184
• Logits/rejected: -0.9556
• Logits/chosen: -0.8877

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: 1e-07
• train_batch_size: 2
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 4
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 100
• training_steps: 250

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.5536	0.2667	50	0.4820	0.0693	-0.4185	1.0	0.4879	-42.8969	-24.6590	-0.9527	-0.8859
0.0373	0.5333	100	0.0199	0.3933	-3.6217	1.0	4.0149	-53.5739	-23.5793	-0.9535	-0.8872
0.0006	0.8	150	0.0004	1.1081	-6.9246	1.0	8.0327	-64.5837	-21.1964	-0.9548	-0.8873
0.0003	1.0667	200	0.0003	1.1629	-7.2750	1.0	8.4380	-65.7519	-21.0137	-0.9548	-0.8871
0.0003	1.3333	250	0.0003	1.1615	-7.2781	1.0	8.4396	-65.7620	-21.0184	-0.9556	-0.8877

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.0+cu121
• Datasets 2.21.0
• Tokenizers 0.19.1