Llama-31-8B_task-3_120-samples_config-2_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: 0.3272

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
2.2357	0.9091	5	2.1322
1.6899	2.0	11	1.3005
0.6881	2.9091	16	0.6189
0.5601	4.0	22	0.4413
0.3305	4.9091	27	0.3900
0.3093	6.0	33	0.3525
0.2941	6.9091	38	0.3440
0.2705	8.0	44	0.3357
0.255	8.9091	49	0.3278
0.1954	10.0	55	0.3272
0.1796	10.9091	60	0.3391
0.1892	12.0	66	0.3651
0.1795	12.9091	71	0.3963
0.0982	14.0	77	0.4424
0.0568	14.9091	82	0.5230
0.046	16.0	88	0.6079
0.0156	16.9091	93	0.6626

Framework versions

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