llama3-8b-instruct-qlora-mini

This model is a fine-tuned version of LoftQ/Meta-Llama-3-8B-Instruct-4bit-64rank on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.8668

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: 2e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss
2.3909	1.0	53	1.5473
2.1937	2.0	106	1.2690
2.0915	3.0	159	1.0977
1.9927	4.0	212	1.0320
1.9058	5.0	265	1.0046
1.8032	6.0	318	0.9885
1.6688	7.0	371	0.9754
1.5215	8.0	424	0.9745
1.3617	9.0	477	0.9640
1.2074	10.0	530	0.9579
1.0429	11.0	583	0.9441
0.9013	12.0	636	0.9355
0.7969	13.0	689	0.9278
0.7092	14.0	742	0.9171
0.6272	15.0	795	0.9070
0.5688	16.0	848	0.9052
0.5128	17.0	901	0.8942
0.469	18.0	954	0.8894
0.4294	19.0	1007	0.8871
0.3953	20.0	1060	0.8807
0.371	21.0	1113	0.8756
0.3533	22.0	1166	0.8750
0.3335	23.0	1219	0.8730
0.3212	24.0	1272	0.8699
0.3108	25.0	1325	0.8687
0.3089	26.0	1378	0.8676
0.3031	27.0	1431	0.8678
0.3014	28.0	1484	0.8675
0.3013	29.0	1537	0.8666
0.2978	30.0	1590	0.8668

Framework versions

• PEFT 0.10.0
• Transformers 4.40.0
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.19.1