Llama2-7B-lora-r-32-generic-step-1200-lr-1e-5-labels_40.0

This model is a fine-tuned version of meta-llama/Llama-2-7b-hf on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 2.7836

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-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 20
• training_steps: 1200
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
5.5017	0.3653	20	5.4071
4.8201	0.7306	40	4.6726
4.2999	1.0959	60	4.1756
3.9694	1.4612	80	3.9211
3.7953	1.8265	100	3.7403
3.5871	2.1918	120	3.6054
3.5058	2.5571	140	3.4953
3.4235	2.9224	160	3.4057
3.3455	3.2877	180	3.3265
3.339	3.6530	200	3.2594
3.1913	4.0183	220	3.2011
3.1476	4.3836	240	3.1500
3.0893	4.7489	260	3.1085
3.0121	5.1142	280	3.0687
2.9902	5.4795	300	3.0374
2.9737	5.8447	320	3.0058
2.8389	6.2100	340	2.9831
2.9038	6.5753	360	2.9586
2.818	6.9406	380	2.9354
2.8141	7.3059	400	2.9206
2.7438	7.6712	420	2.9014
2.7746	8.0365	440	2.8862
2.7449	8.4018	460	2.8748
2.738	8.7671	480	2.8616
2.6404	9.1324	500	2.8548
2.6821	9.4977	520	2.8463
2.6512	9.8630	540	2.8365
2.6048	10.2283	560	2.8332
2.6179	10.5936	580	2.8247
2.6451	10.9589	600	2.8168
2.5769	11.3242	620	2.8157
2.6271	11.6895	640	2.8098
2.6242	12.0548	660	2.8061
2.5967	12.4201	680	2.8062
2.5563	12.7854	700	2.7994
2.4609	13.1507	720	2.7982
2.5047	13.5160	740	2.7984
2.5078	13.8813	760	2.7911
2.5034	14.2466	780	2.7917
2.5125	14.6119	800	2.7891
2.4899	14.9772	820	2.7882
2.4671	15.3425	840	2.7891
2.4907	15.7078	860	2.7873
2.5025	16.0731	880	2.7852
2.4492	16.4384	900	2.7864
2.761	16.8037	920	2.7858
2.4556	17.1689	940	2.7847
2.4823	17.5342	960	2.7845
2.4832	17.8995	980	2.7833
2.4917	18.2648	1000	2.7838
2.4228	18.6301	1020	2.7838
2.7161	18.9954	1040	2.7829
2.4172	19.3607	1060	2.7828
2.7152	19.7260	1080	2.7830
2.4624	20.0913	1100	2.7832
2.4348	20.4566	1120	2.7833
2.4513	20.8219	1140	2.7835
2.4233	21.1872	1160	2.7835
2.4601	21.5525	1180	2.7836
2.4652	21.9178	1200	2.7836

Framework versions

• PEFT 0.15.2
• Transformers 4.45.2
• Pytorch 2.5.0+cu121
• Datasets 3.2.0
• Tokenizers 0.20.3