Llama-31-8B_task-1_60-samples_config-3

This model is a fine-tuned version of meta-llama/Meta-Llama-3.1-8B-Instruct on the GaetanMichelet/chat-60_ft_task-1 dataset. It achieves the following results on the evaluation set:

• Loss: 1.2924

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: 1
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• gradient_accumulation_steps: 8
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 150

Training results

Training Loss	Epoch	Step	Validation Loss
2.1691	0.8696	5	2.0741
2.0145	1.9130	11	2.0674
2.0918	2.9565	17	2.0492
2.0838	4.0	23	2.0192
2.0314	4.8696	28	1.9792
1.9775	5.9130	34	1.9190
1.8873	6.9565	40	1.8339
1.7547	8.0	46	1.7314
1.6653	8.8696	51	1.6435
1.5709	9.9130	57	1.5691
1.533	10.9565	63	1.5254
1.4035	12.0	69	1.4860
1.4227	12.8696	74	1.4573
1.4167	13.9130	80	1.4216
1.3733	14.9565	86	1.3884
1.2917	16.0	92	1.3621
1.2393	16.8696	97	1.3432
1.1512	17.9130	103	1.3246
1.1361	18.9565	109	1.3081
1.089	20.0	115	1.2985
1.0272	20.8696	120	1.2924
1.0591	21.9130	126	1.2934
0.9601	22.9565	132	1.3023
0.9245	24.0	138	1.3152
0.8188	24.8696	143	1.3258
0.8866	25.9130	149	1.3491
0.7508	26.9565	155	1.3779
0.7961	28.0	161	1.4176

Framework versions

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