Llama2-7B-lora-r-32-generic-step-800-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: 2.8780

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: 800
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
5.6038	0.3653	20	5.3771
4.8695	0.7306	40	4.6843
4.3239	1.0959	60	4.2063
4.0079	1.4612	80	3.9576
3.8164	1.8265	100	3.7695
3.5871	2.1918	120	3.6264
3.5886	2.5571	140	3.5123
3.4651	2.9224	160	3.4222
3.3567	3.2877	180	3.3467
3.3554	3.6530	200	3.2796
3.1688	4.0183	220	3.2285
3.1091	4.3836	240	3.1811
3.1209	4.7489	260	3.1415
3.037	5.1142	280	3.1061
3.0502	5.4795	300	3.0724
2.9951	5.8447	320	3.0460
2.9993	6.2100	340	3.0218
2.9732	6.5753	360	3.0007
2.9087	6.9406	380	2.9797
2.8941	7.3059	400	2.9649
2.8462	7.6712	420	2.9497
2.8651	8.0365	440	2.9379
2.8268	8.4018	460	2.9282
2.761	8.7671	480	2.9175
2.7696	9.1324	500	2.9100
2.7591	9.4977	520	2.9049
2.7767	9.8630	540	2.8982
2.757	10.2283	560	2.8941
2.7148	10.5936	580	2.8897
2.7326	10.9589	600	2.8876
2.7234	11.3242	620	2.8851
2.8103	11.6895	640	2.8834
2.7953	12.0548	660	2.8811
2.7073	12.4201	680	2.8802
2.7316	12.7854	700	2.8794
2.704	13.1507	720	2.8785
2.6919	13.5160	740	2.8783
2.7829	13.8813	760	2.8781
2.719	14.2466	780	2.8780
2.6766	14.6119	800	2.8780

Framework versions

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