Llama2-7B-lora-r-32-generic-step-1400-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: 2.9031

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: 1400
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
5.4792	0.3653	20	5.4355
4.846	0.7306	40	4.7021
4.2576	1.0959	60	4.1844
3.9857	1.4612	80	3.9113
3.8782	1.8265	100	3.7408
3.7144	2.1918	120	3.6063
3.5158	2.5571	140	3.4962
3.4588	2.9224	160	3.4011
3.3361	3.2877	180	3.3260
3.2559	3.6530	200	3.2620
3.2774	4.0183	220	3.2042
3.1416	4.3836	240	3.1538
3.0975	4.7489	260	3.1082
3.0383	5.1142	280	3.0682
3.0127	5.4795	300	3.0342
2.9291	5.8447	320	2.9985
2.9167	6.2100	340	2.9765
2.8798	6.5753	360	2.9489
2.8393	6.9406	380	2.9254
2.8357	7.3059	400	2.9115
2.8181	7.6712	420	2.8913
2.7272	8.0365	440	2.8738
2.6862	8.4018	460	2.8638
2.7263	8.7671	480	2.8548
2.6714	9.1324	500	2.8431
2.9486	9.4977	520	2.8338
2.6906	9.8630	540	2.8236
2.575	10.2283	560	2.8209
2.5791	10.5936	580	2.8145
2.5564	10.9589	600	2.8053
2.5934	11.3242	620	2.8057
2.5011	11.6895	640	2.7961
2.4966	12.0548	660	2.7931
2.5523	12.4201	680	2.7949
2.4702	12.7854	700	2.7896
2.4779	13.1507	720	2.7885
2.4192	13.5160	740	2.7853
2.4555	13.8813	760	2.7842
2.6024	14.2466	780	2.7853
2.4411	14.6119	800	2.7838
2.467	14.9772	820	2.7815
2.395	15.3425	840	2.7825
2.4051	15.7078	860	2.7838
2.4257	16.0731	880	2.7809
2.4494	16.4384	900	2.7858
2.3651	16.8037	920	2.7840
2.3755	17.1689	940	2.7784
2.7721	17.5342	960	2.7835
2.5286	17.8995	980	2.7902
2.4642	18.2648	1000	2.8004
2.7115	18.6301	1020	2.8042
2.4904	18.9954	1040	2.8173
2.5253	19.3607	1060	2.8436
2.5782	19.7260	1080	2.8669
2.573	20.0913	1100	2.8835
2.5573	20.4566	1120	2.8946
2.8642	20.8219	1140	2.9031

Framework versions

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