Llama-31-8B_task-3_120-samples_config-1_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 and the GaetanMichelet/chat-120_ft_task-3_auto datasets. It achieves the following results on the evaluation set:

• Loss: 1.1182

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: 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: 50

Training results

Training Loss	Epoch	Step	Validation Loss
1.6493	1.0	11	1.6402
1.4499	2.0	22	1.4987
1.3516	3.0	33	1.3295
1.2309	4.0	44	1.1897
1.151	5.0	55	1.1514
1.1003	6.0	66	1.1307
1.0537	7.0	77	1.1200
1.0287	8.0	88	1.1182
0.9295	9.0	99	1.1300
0.8133	10.0	110	1.1687
0.8071	11.0	121	1.2087
0.6155	12.0	132	1.2670
0.6191	13.0	143	1.2891
0.3871	14.0	154	1.4231
0.2973	15.0	165	1.3964

Framework versions

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