Llama-31-8B_task-3_180-samples_config-2_full_auto

This model is a fine-tuned version of meta-llama/Meta-Llama-3.1-8B-Instruct on the GaetanMichelet/chat-60_ft_task-3_auto, the GaetanMichelet/chat-120_ft_task-3_auto and the GaetanMichelet/chat-180_ft_task-3_auto datasets. It achieves the following results on the evaluation set:

• Loss: 1.1260

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
1.674	0.9412	8	1.6584
1.5596	2.0	17	1.5302
1.4536	2.9412	25	1.4046
1.1955	4.0	34	1.2397
1.1221	4.9412	42	1.1855
1.1181	6.0	51	1.1592
1.1246	6.9412	59	1.1434
1.0507	8.0	68	1.1335
0.9763	8.9412	76	1.1265
0.9776	10.0	85	1.1260
0.929	10.9412	93	1.1289
0.8679	12.0	102	1.1548
0.8616	12.9412	110	1.1690
0.815	14.0	119	1.2061
0.6912	14.9412	127	1.2300
0.6046	16.0	136	1.2953
0.5919	16.9412	144	1.3360

Framework versions

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