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

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: 1.1336

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
1.7269	0.9091	5	1.6901
1.6699	2.0	11	1.5939
1.5273	2.9091	16	1.5005
1.3953	4.0	22	1.3805
1.2642	4.9091	27	1.2621
1.1571	6.0	33	1.1986
1.1345	6.9091	38	1.1788
1.1145	8.0	44	1.1606
1.0654	8.9091	49	1.1505
1.0821	10.0	55	1.1417
1.0346	10.9091	60	1.1381
1.0027	12.0	66	1.1336
0.9693	12.9091	71	1.1350
0.9594	14.0	77	1.1414
0.8512	14.9091	82	1.1548
0.863	16.0	88	1.1607
0.8028	16.9091	93	1.1746
0.731	18.0	99	1.2033
0.7089	18.9091	104	1.2296

Framework versions

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