Llama-31-8B_task-3_120-samples_config-2

This model is a fine-tuned version of meta-llama/Meta-Llama-3.1-8B-Instruct on the GaetanMichelet/chat-60_ft_task-3 and the GaetanMichelet/chat-120_ft_task-3 datasets. It achieves the following results on the evaluation set:

• Loss: 0.4302

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: 0.0001
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• gradient_accumulation_steps: 16
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss
2.4469	0.9091	5	2.3539
1.8346	2.0	11	1.4922
0.7576	2.9091	16	0.7652
0.6409	4.0	22	0.5627
0.4304	4.9091	27	0.5238
0.3624	6.0	33	0.4705
0.3967	6.9091	38	0.4452
0.3293	8.0	44	0.4328
0.2432	8.9091	49	0.4302
0.2102	10.0	55	0.4359
0.2004	10.9091	60	0.4583
0.1634	12.0	66	0.4724
0.1177	12.9091	71	0.5530
0.0376	14.0	77	0.7361
0.0204	14.9091	82	0.7768
0.0118	16.0	88	0.8608

Framework versions

• PEFT 0.12.0
• Transformers 4.44.0
• Pytorch 2.1.2+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1