Llama-31-8B_task-2_60-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-2_auto dataset. It achieves the following results on the evaluation set:

• Loss: 0.5986

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
0.8574	0.6957	2	0.8904
0.8746	1.7391	5	0.8490
0.7842	2.7826	8	0.7717
0.728	3.8261	11	0.7140
0.6405	4.8696	14	0.6587
0.5991	5.9130	17	0.6365
0.5371	6.9565	20	0.6147
0.5007	8.0	23	0.6018
0.4783	8.6957	25	0.5986
0.4118	9.7391	28	0.6050
0.4002	10.7826	31	0.6278
0.3234	11.8261	34	0.6697
0.2816	12.8696	37	0.7073
0.2159	13.9130	40	0.7197
0.1785	14.9565	43	0.7951
0.1209	16.0	46	0.9001

Framework versions

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