exceptions_exp2_swap_0.7_cost_to_drop_5039

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

• Loss: 3.5645
• Accuracy: 0.3686

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: 0.0006
• train_batch_size: 16
• eval_batch_size: 16
• seed: 5039
• gradient_accumulation_steps: 5
• total_train_batch_size: 80
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.98) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• num_epochs: 50.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
4.8235	0.2917	1000	4.7495	0.2552
4.3462	0.5834	2000	4.2916	0.2988
4.1645	0.8750	3000	4.1072	0.3144
4.0081	1.1665	4000	3.9962	0.3242
3.9471	1.4582	5000	3.9234	0.3309
3.8937	1.7499	6000	3.8664	0.3353
3.7596	2.0414	7000	3.8213	0.3403
3.762	2.3331	8000	3.7930	0.3430
3.7331	2.6248	9000	3.7621	0.3462
3.7349	2.9165	10000	3.7355	0.3487
3.6362	3.2080	11000	3.7221	0.3505
3.6457	3.4996	12000	3.7049	0.3522
3.6514	3.7913	13000	3.6858	0.3539
3.5526	4.0828	14000	3.6795	0.3549
3.5965	4.3745	15000	3.6687	0.3559
3.5821	4.6662	16000	3.6562	0.3573
3.579	4.9579	17000	3.6405	0.3585
3.5144	5.2494	18000	3.6428	0.3592
3.5321	5.5411	19000	3.6311	0.3600
3.5373	5.8327	20000	3.6227	0.3609
3.4516	6.1243	21000	3.6268	0.3612
3.4813	6.4159	22000	3.6168	0.3618
3.4953	6.7076	23000	3.6082	0.3627
3.4995	6.9993	24000	3.5997	0.3632
3.4379	7.2908	25000	3.6082	0.3634
3.4676	7.5825	26000	3.5976	0.3641
3.4706	7.8742	27000	3.5889	0.3651
3.3905	8.1657	28000	3.5985	0.3648
3.4272	8.4574	29000	3.5909	0.3652
3.4349	8.7490	30000	3.5824	0.3661
3.3349	9.0405	31000	3.5871	0.3661
3.3855	9.3322	32000	3.5864	0.3661
3.4069	9.6239	33000	3.5773	0.3665
3.429	9.9156	34000	3.5690	0.3676
3.3396	10.2071	35000	3.5845	0.3668
3.3802	10.4988	36000	3.5740	0.3674
3.3967	10.7905	37000	3.5667	0.3680
3.3041	11.0820	38000	3.5735	0.3681
3.3322	11.3736	39000	3.5746	0.3679
3.3694	11.6653	40000	3.5645	0.3686
3.3702	11.9570	41000	3.5579	0.3689
3.3186	12.2485	42000	3.5710	0.3687
3.3503	12.5402	43000	3.5659	0.3690
3.3611	12.8319	44000	3.5579	0.3696
3.2771	13.1234	45000	3.5718	0.3690
3.3066	13.4151	46000	3.5658	0.3692
3.3455	13.7067	47000	3.5570	0.3700
3.3451	13.9984	48000	3.5483	0.3702
3.2904	14.2899	49000	3.5660	0.3694
3.3167	14.5816	50000	3.5563	0.3700
3.3306	14.8733	51000	3.5501	0.3707
3.2528	15.1648	52000	3.5641	0.3699
3.2853	15.4565	53000	3.5608	0.3704
3.3115	15.7482	54000	3.5489	0.3711
3.2167	16.0397	55000	3.5650	0.3704
3.2616	16.3313	56000	3.5638	0.3707
3.279	16.6230	57000	3.5539	0.3711
3.3055	16.9147	58000	3.5442	0.3714
3.2311	17.2062	59000	3.5607	0.3706
3.2687	17.4979	60000	3.5537	0.3712
3.2897	17.7896	61000	3.5465	0.3717
3.1981	18.0811	62000	3.5624	0.3711
3.2564	18.3728	63000	3.5579	0.3714
3.2645	18.6644	64000	3.5481	0.3715
3.2763	18.9561	65000	3.5416	0.3723
3.2251	19.2476	66000	3.5598	0.3715
3.2513	19.5393	67000	3.5499	0.3719
3.2806	19.8310	68000	3.5451	0.3723
3.1922	20.1225	69000	3.5614	0.3715
3.226	20.4142	70000	3.5556	0.3719
3.2528	20.7059	71000	3.5446	0.3723
3.2603	20.9975	72000	3.5390	0.3727
3.207	21.2891	73000	3.5593	0.3719
3.2368	21.5807	74000	3.5521	0.3722
3.2494	21.8724	75000	3.5417	0.3730
3.1753	22.1639	76000	3.5601	0.3718
3.2161	22.4556	77000	3.5512	0.3725
3.227	22.7473	78000	3.5419	0.3729
3.1477	23.0388	79000	3.5599	0.3723
3.1914	23.3305	80000	3.5553	0.3725
3.2168	23.6222	81000	3.5480	0.3728
3.2271	23.9138	82000	3.5429	0.3731
3.1566	24.2053	83000	3.5582	0.3724
3.2024	24.4970	84000	3.5524	0.3726
3.2178	24.7887	85000	3.5462	0.3732
3.1374	25.0802	86000	3.5614	0.3724
3.1895	25.3719	87000	3.5549	0.3729
3.1943	25.6636	88000	3.5485	0.3732
3.2212	25.9553	89000	3.5405	0.3736
3.1513	26.2468	90000	3.5602	0.3727
3.1832	26.5384	91000	3.5531	0.3733
3.2104	26.8301	92000	3.5431	0.3737

Framework versions

• Transformers 4.55.2
• Pytorch 2.8.0+cu128
• Datasets 4.0.0
• Tokenizers 0.21.4