Llama-31-8B_task-2_180-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-2, the GaetanMichelet/chat-120_ft_task-2 and the GaetanMichelet/chat-180_ft_task-2 datasets. It achieves the following results on the evaluation set:

• Loss: 0.7009

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
1.0575	0.9412	8	1.0786
0.9283	2.0	17	0.8986
0.8658	2.9412	25	0.8192
0.7202	4.0	34	0.7616
0.6781	4.9412	42	0.7277
0.648	6.0	51	0.7058
0.5956	6.9412	59	0.7009
0.5123	8.0	68	0.7151
0.3458	8.9412	76	0.8068
0.2817	10.0	85	0.8643
0.1755	10.9412	93	1.0694
0.0778	12.0	102	1.1664
0.0487	12.9412	110	1.3312
0.028	14.0	119	1.5136

Framework versions

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