Llama-3.2-3B-Instruct-arkey_emails-qlora

This model is a fine-tuned version of meta-llama/Llama-3.2-3B-Instruct on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 2.4612

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: 5e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 64
• total_train_batch_size: 512
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss
No log	1.0	2	2.9993
No log	2.0	4	2.9668
No log	3.0	6	2.9307
No log	4.0	8	2.8914
No log	5.0	10	2.8501
No log	6.0	12	2.8096
No log	7.0	14	2.7724
No log	8.0	16	2.7370
No log	9.0	18	2.7032
No log	10.0	20	2.6715
No log	11.0	22	2.6420
No log	12.0	24	2.6153
No log	13.0	26	2.5913
No log	14.0	28	2.5698
No log	15.0	30	2.5506
No log	16.0	32	2.5336
No log	17.0	34	2.5188
No log	18.0	36	2.5058
No log	19.0	38	2.4944
No log	20.0	40	2.4848
No log	21.0	42	2.4767
No log	22.0	44	2.4703
No log	23.0	46	2.4656
No log	24.0	48	2.4625
No log	25.0	50	2.4612

Framework versions

• PEFT 0.15.2
• Transformers 4.51.3
• Pytorch 2.1.2
• Datasets 3.5.1
• Tokenizers 0.21.1