Na_M2_1000steps_1e8rate_03beta_cSFTDPO

This model is a fine-tuned version of tsavage68/Na_M2_1000steps_1e7_SFT on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.4450
• Rewards/chosen: 0.1680
• Rewards/rejected: -0.4255
• Rewards/accuracies: 1.0
• Rewards/margins: 0.5934
• Logps/rejected: -81.3416
• Logps/chosen: -47.5724
• Logits/rejected: -2.5355
• Logits/chosen: -2.5481

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-08
• 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: 1000

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.6955	0.2667	50	0.6882	0.0099	-0.0031	0.5600	0.0130	-79.9338	-48.0995	-2.5354	-2.5481
0.6761	0.5333	100	0.6730	0.0130	-0.0315	0.6600	0.0445	-80.0283	-48.0889	-2.5363	-2.5489
0.6154	0.8	150	0.5971	0.0672	-0.1393	0.9800	0.2065	-80.3878	-47.9083	-2.5367	-2.5493
0.5735	1.0667	200	0.5430	0.1029	-0.2302	1.0	0.3331	-80.6906	-47.7893	-2.5352	-2.5478
0.5047	1.3333	250	0.5020	0.1363	-0.3030	1.0	0.4393	-80.9334	-47.6779	-2.5353	-2.5478
0.4525	1.6	300	0.4751	0.1411	-0.3685	1.0	0.5096	-81.1517	-47.6622	-2.5350	-2.5476
0.451	1.8667	350	0.4572	0.1576	-0.3988	1.0	0.5564	-81.2528	-47.6072	-2.5350	-2.5475
0.4434	2.1333	400	0.4501	0.1391	-0.4387	1.0	0.5778	-81.3857	-47.6686	-2.5351	-2.5477
0.4313	2.4	450	0.4454	0.1528	-0.4370	1.0	0.5899	-81.3802	-47.6230	-2.5343	-2.5469
0.4546	2.6667	500	0.4513	0.1462	-0.4293	1.0	0.5755	-81.3544	-47.6450	-2.5345	-2.5471
0.4526	2.9333	550	0.4424	0.1917	-0.4110	1.0	0.6027	-81.2934	-47.4934	-2.5352	-2.5476
0.4426	3.2	600	0.4437	0.1805	-0.4175	1.0	0.5980	-81.3150	-47.5307	-2.5361	-2.5486
0.4452	3.4667	650	0.4403	0.1651	-0.4392	1.0	0.6043	-81.3875	-47.5821	-2.5347	-2.5473
0.418	3.7333	700	0.4450	0.1668	-0.4237	1.0	0.5905	-81.3358	-47.5764	-2.5348	-2.5474
0.4281	4.0	750	0.4450	0.1680	-0.4255	1.0	0.5934	-81.3416	-47.5724	-2.5355	-2.5481
0.4503	4.2667	800	0.4450	0.1680	-0.4255	1.0	0.5934	-81.3416	-47.5724	-2.5355	-2.5481
0.4372	4.5333	850	0.4450	0.1680	-0.4255	1.0	0.5934	-81.3416	-47.5724	-2.5355	-2.5481
0.4135	4.8	900	0.4450	0.1680	-0.4255	1.0	0.5934	-81.3416	-47.5724	-2.5355	-2.5481
0.4316	5.0667	950	0.4450	0.1680	-0.4255	1.0	0.5934	-81.3416	-47.5724	-2.5355	-2.5481
0.4438	5.3333	1000	0.4450	0.1680	-0.4255	1.0	0.5934	-81.3416	-47.5724	-2.5355	-2.5481

Framework versions

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