Llama2-7B-lora-r-32-generic-step-1800-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: 2.8495

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: 1800
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
5.5122	0.3653	20	5.5248
4.8114	0.7306	40	4.8096
4.2874	1.0959	60	4.2869
4.0767	1.4612	80	4.0173
3.8146	1.8265	100	3.8281
3.6484	2.1918	120	3.6863
3.4968	2.5571	140	3.5713
3.449	2.9224	160	3.4708
3.3535	3.2877	180	3.3962
3.2769	3.6530	200	3.3254
3.1924	4.0183	220	3.2663
3.1511	4.3836	240	3.2146
3.043	4.7489	260	3.1650
3.0218	5.1142	280	3.1280
2.9589	5.4795	300	3.0957
2.9661	5.8447	320	3.0585
2.8736	6.2100	340	3.0300
2.8571	6.5753	360	3.0027
2.8029	6.9406	380	2.9830
2.7944	7.3059	400	2.9638
2.7186	7.6712	420	2.9465
2.7403	8.0365	440	2.9255
2.6509	8.4018	460	2.9177
2.664	8.7671	480	2.9052
2.6223	9.1324	500	2.8964
2.6088	9.4977	520	2.8872
2.6151	9.8630	540	2.8811
2.6091	10.2283	560	2.8776
2.5445	10.5936	580	2.8636
2.5474	10.9589	600	2.8582
2.7061	11.3242	620	2.8577
2.5042	11.6895	640	2.8481
2.4626	12.0548	660	2.8455
2.462	12.4201	680	2.8460
2.4464	12.7854	700	2.8432
2.4624	13.1507	720	2.8450
2.3736	13.5160	740	2.8409
2.4384	13.8813	760	2.8398
2.3796	14.2466	780	2.8407
2.334	14.6119	800	2.8389
2.3981	14.9772	820	2.8309
2.3528	15.3425	840	2.8420
2.2988	15.7078	860	2.8335
2.3556	16.0731	880	2.8346
2.2997	16.4384	900	2.8415
2.3096	16.8037	920	2.8390
2.5605	17.1689	940	2.8427
2.2552	17.5342	960	2.8439
2.2732	17.8995	980	2.8453
2.2488	18.2648	1000	2.8533
2.2665	18.6301	1020	2.8495

Framework versions

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