Llama2-7B-lora-r-32-generic-step-600-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: 3.0888

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: 600
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
5.5848	0.3653	20	5.4961
4.8149	0.7306	40	4.7747
4.302	1.0959	60	4.2724
4.0321	1.4612	80	4.0219
3.7948	1.8265	100	3.8366
3.591	2.1918	120	3.7064
3.5765	2.5571	140	3.5995
3.4451	2.9224	160	3.5143
3.3793	3.2877	180	3.4442
3.33	3.6530	200	3.3844
3.2783	4.0183	220	3.3372
3.1619	4.3836	240	3.2939
3.1992	4.7489	260	3.2555
3.1576	5.1142	280	3.2251
3.0138	5.4795	300	3.1994
3.018	5.8447	320	3.1777
3.0246	6.2100	340	3.1610
3.0205	6.5753	360	3.1449
2.9761	6.9406	380	3.1306
3.0076	7.3059	400	3.1208
2.9228	7.6712	420	3.1121
2.9606	8.0365	440	3.1050
2.9182	8.4018	460	3.0997
2.9403	8.7671	480	3.0960
2.9776	9.1324	500	3.0929
2.9118	9.4977	520	3.0910
2.9364	9.8630	540	3.0897
2.9185	10.2283	560	3.0890
2.9412	10.5936	580	3.0888
2.8851	10.9589	600	3.0888

Framework versions

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