Llama-31-8B_task-1_60-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-1_auto dataset. It achieves the following results on the evaluation set:

• Loss: 0.8247

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
2.1899	0.8696	5	2.0616
1.9473	1.9130	11	1.7736
1.6049	2.9565	17	1.4599
1.1934	4.0	23	1.0350
0.8711	4.8696	28	0.9062
0.8035	5.9130	34	0.8637
0.7366	6.9565	40	0.8406
0.6995	8.0	46	0.8247
0.6613	8.8696	51	0.8259
0.5531	9.9130	57	0.8318
0.5061	10.9565	63	0.8598
0.3776	12.0	69	0.9167
0.264	12.8696	74	1.0158
0.1878	13.9130	80	1.1250
0.1417	14.9565	86	1.2038

Framework versions

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