Llama3.1-8b-instruct-SFT-2024-09-19_LoRAs_r32

This model is a fine-tuned version of meta-llama/Meta-Llama-3.1-8B-Instruct on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.1544

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: 6
• eval_batch_size: 1
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant
• num_epochs: 1.5

Training results

Training Loss	Epoch	Step	Validation Loss
1.571	0.0793	1000	1.4287
1.3921	0.1586	2000	1.3664
1.3336	0.2380	3000	1.3296
1.2962	0.3173	4000	1.3049
1.2721	0.3966	5000	1.2862
1.2481	0.4759	6000	1.2782
1.2365	0.5552	7000	1.2647
1.2117	0.6346	8000	1.2509
1.1997	0.7139	9000	1.2259
1.197	0.7932	10000	1.2185
1.1707	0.8725	11000	1.2055
1.165	0.9519	12000	1.1917
1.1348	1.0312	13000	1.1837
1.1124	1.1105	14000	1.1803
1.1063	1.1898	15000	1.1705
1.0954	1.2691	16000	1.1635
1.087	1.3485	17000	1.1581
1.0834	1.4278	18000	1.1544

Framework versions

• PEFT 0.12.0
• Transformers 4.44.2
• Pytorch 2.0.1+cu118
• Datasets 3.0.0
• Tokenizers 0.19.1